In order to automate banking processes (e.g. payments, money transfers, foreign trade), we need to extract banking transactions from different types of mediums such as faxes, e-mails, and scanners. Banking orders may be considered as complex documents since they contain quite complex relations compared to traditional datasets used in relation extraction research. In this paper, we present our method to extract intersentential, nested and complex relations from banking orders, and introduce a relation extraction method based on maximal clique factorization technique. We demonstrate 11% error reduction over previous methods.
Extraction of financial and economic events from text has previously been done mostly using rule-based methods, with more recent works employing machine learning techniques. This work is in line with this latter approach, leveraging relevant Wikipedia sections to extract weak labels for sentences describing economic events. Whereas previous weakly supervised approaches required a knowledge-base of such events, or corresponding financial figures, our approach requires no such additional data, and can be employed to extract economic events related to companies which are not even mentioned in the training data.
We examine the affective content of central bank press statements using emotion analysis. Our focus is on two major international players, the European Central Bank (ECB) and the US Federal Reserve Bank (Fed), covering a time span from 1998 through 2019. We reveal characteristic patterns in the emotional dimensions of valence, arousal, and dominance and find—despite the commonly established attitude that emotional wording in central bank communication should be avoided—a correlation between the state of the economy and particularly the dominance dimension in the press releases under scrutiny and, overall, an impact of the president in office.
In this paper, we show deep learning models can be used to forecast firm material event sequences based on the contents in the company’s 8-K Current Reports. Specifically, we exploit state-of-the-art neural architectures, including sequence-to-sequence (Seq2Seq) architecture and attention mechanisms, in the model. Our 8K-powered deep learning model demonstrates promising performance in forecasting firm future event sequences. The model is poised to benefit various stakeholders, including management and investors, by facilitating risk management and decision making.
Considering event structure information has proven helpful in text-based stock movement prediction. However, existing works mainly adopt the coarse-grained events, which loses the specific semantic information of diverse event types. In this work, we propose to incorporate the fine-grained events in stock movement prediction. Firstly, we propose a professional finance event dictionary built by domain experts and use it to extract fine-grained events automatically from finance news. Then we design a neural model to combine finance news with fine-grained event structure and stock trade data to predict the stock movement. Besides, in order to improve the generalizability of the proposed method, we design an advanced model that uses the extracted fine-grained events as the distant supervised label to train a multi-task framework of event extraction and stock prediction. The experimental results show that our method outperforms all the baselines and has good generalizability.
Incorporating related text information has proven successful in stock market prediction. However, it is a huge challenge to utilize texts in the enormous forex (foreign currency exchange) market because the associated texts are too redundant. In this work, we propose a BERT-based Hierarchical Aggregation Model to summarize a large amount of finance news to predict forex movement. We firstly group news from different aspects: time, topic and category. Then we extract the most crucial news in each group by the SOTA extractive summarization method. Finally, we conduct interaction between the news and the trade data with attention to predict the forex movement. The experimental results show that the category based method performs best among three grouping methods and outperforms all the baselines. Besides, we study the influence of essential news attributes (category and region) by statistical analysis and summarize the influence patterns for different currency pairs.
Governmental institutions are employing artificial intelligence techniques to deal with their specific problems and exploit their huge amounts of both structured and unstructured information. In particular, natural language processing and machine learning techniques are being used to process citizen feedback. In this paper, we report on the use of such techniques for analyzing and classifying complaints, in the context of the Portuguese Economic and Food Safety Authority. Grounded in its operational process, we address three different classification problems: target economic activity, implied infraction severity level, and institutional competence. We show promising results obtained using feature-based approaches and traditional classifiers, with accuracy scores above 70%, and analyze the shortcomings of our current results and avenues for further improvement, taking into account the intended use of our classifiers in helping human officers to cope with thousands of yearly complaints.
With conversational agents or chatbots making up in quantity of replies rather than quality, the need to identify user intent has become a main concern to improve these agents. Dialog act (DA) classification tackles this concern, and while existing studies have already addressed DA classification in general contexts, no training corpora in the context of e-commerce is available to the public. This research addressed the said insufficiency by building a text-based corpus of 7,265 posts from the question and answer section of products on Lazada Philippines. The SWBD-DAMSL tagset for DA classification was modified to 28 tags fitting the categories applicable to e-commerce conversations. The posts were annotated manually by three (3) human annotators and preprocessing techniques decreased the vocabulary size from 6,340 to 1,134. After analysis, the corpus was composed dominantly of single-label posts, with 34% of the corpus having multiple intent tags. The annotated corpus allowed insights toward the structure of posts created with single to multiple intents.
In recent years, the task of Question Answering over passages, also pitched as a reading comprehension, has evolved into a very active research area. A reading comprehension system extracts a span of text, comprising of named entities, dates, small phrases, etc., which serve as the answer to a given question. However, these spans of text would result in an unnatural reading experience in a conversational system. Usually, dialogue systems solve this issue by using template-based language generation. These systems, though adequate for a domain specific task, are too restrictive and predefined for a domain independent system. In order to present the user with a more conversational experience, we propose a pointer generator based full-length answer generator which can be used with most QA systems. Our system generates a full length answer given a question and the extracted factoid/span answer without relying on the passage from where the answer was extracted. We also present a dataset of 315000 question, factoid answer and full length answer triples. We have evaluated our system using ROUGE-1,2,L and BLEU and achieved 74.05 BLEU score and 86.25 Rogue-L score.
As an attempt to combine extractive and abstractive summarization, Sentence Rewriting models adopt the strategy of extracting salient sentences from a document first and then paraphrasing the selected ones to generate a summary. However, the existing models in this framework mostly rely on sentence-level rewards or suboptimal labels, causing a mismatch between a training objective and evaluation metric. In this paper, we present a novel training signal that directly maximizes summary-level ROUGE scores through reinforcement learning. In addition, we incorporate BERT into our model, making good use of its ability on natural language understanding. In extensive experiments, we show that a combination of our proposed model and training procedure obtains new state-of-the-art performance on both CNN/Daily Mail and New York Times datasets. We also demonstrate that it generalizes better on DUC-2002 test set.
Timeline summarization (TLS) automatically identifies key dates of major events and provides short descriptions of what happened on these dates. Previous approaches to TLS have focused on extractive methods. In contrast, we suggest an abstractive timeline summarization system. Our system is entirely unsupervised, which makes it especially suited to TLS where there are very few gold summaries available for training of supervised systems. In addition, we present the first abstractive oracle experiments for TLS. Our system outperforms extractive competitors in terms of ROUGE when the number of input documents is high and the output requires strong compression. In these cases, our oracle experiments confirm that our approach also has a higher upper bound for ROUGE scores than extractive methods. A study with human judges shows that our abstractive system also produces output that is easy to read and understand.
Linking facts across documents is a challenging task, as the language used to express the same information in a sentence can vary significantly, which complicates the task of multi-document summarization. Consequently, existing approaches heavily rely on hand-crafted features, which are domain-dependent and hard to craft, or additional annotated data, which is costly to gather. To overcome these limitations, we present a novel method, which makes use of two types of sentence embeddings: universal embeddings, which are trained on a large unrelated corpus, and domain-specific embeddings, which are learned during training. To this end, we develop SemSentSum, a fully data-driven model able to leverage both types of sentence embeddings by building a sentence semantic relation graph. SemSentSum achieves competitive results on two types of summary, consisting of 665 bytes and 100 words. Unlike other state-of-the-art models, neither hand-crafted features nor additional annotated data are necessary, and the method is easily adaptable for other tasks. To our knowledge, we are the first to use multiple sentence embeddings for the task of multi-document summarization.
User-generated reviews of products or services provide valuable information to customers. However, it is often impossible to read each of the potentially thousands of reviews: it would therefore save valuable time to provide short summaries of their contents. We address opinion summarization, a multi-document summarization task, with an unsupervised abstractive summarization neural system. Our system is based on (i) a language model that is meant to encode reviews to a vector space, and to generate fluent sentences from the same vector space (ii) a clustering step that groups together reviews about the same aspects and allows the system to generate summary sentences focused on these aspects. Our experiments on the Oposum dataset empirically show the importance of the clustering step.
Automatic summarization methods have been studied on a variety of domains, including news and scientific articles. Yet, legislation has not previously been considered for this task, despite US Congress and state governments releasing tens of thousands of bills every year. In this paper, we introduce BillSum, the first dataset for summarization of US Congressional and California state bills. We explain the properties of the dataset that make it more challenging to process than other domains. Then, we benchmark extractive methods that consider neural sentence representations and traditional contextual features. Finally, we demonstrate that models built on Congressional bills can be used to summarize California billa, thus, showing that methods developed on this dataset can transfer to states without human-written summaries.
We suggest a new idea of Editorial Network – a mixed extractive-abstractive summarization approach, which is applied as a post-processing step over a given sequence of extracted sentences. We further suggest an effective way for training the “editor” based on a novel soft-labeling approach. Using the CNN/DailyMail dataset we demonstrate the effectiveness of our approach compared to state-of-the-art extractive-only or abstractive-only baselines.
Highlighting is a powerful tool to pick out important content and emphasize. Creating summary highlights at the sub-sentence level is particularly desirable, because sub-sentences are more concise than whole sentences. They are also better suited than individual words and phrases that can potentially lead to disfluent, fragmented summaries. In this paper we seek to generate summary highlights by annotating summary-worthy sub-sentences and teaching classifiers to do the same. We frame the task as jointly selecting important sentences and identifying a single most informative textual unit from each sentence. This formulation dramatically reduces the task complexity involved in sentence compression. Our study provides new benchmarks and baselines for generating highlights at the sub-sentence level.
This paper introduces the SAMSum Corpus, a new dataset with abstractive dialogue summaries. We investigate the challenges it poses for automated summarization by testing several models and comparing their results with those obtained on a corpus of news articles. We show that model-generated summaries of dialogues achieve higher ROUGE scores than the model-generated summaries of news – in contrast with human evaluators’ judgement. This suggests that a challenging task of abstractive dialogue summarization requires dedicated models and non-standard quality measures. To our knowledge, our study is the first attempt to introduce a high-quality chat-dialogues corpus, manually annotated with abstractive summarizations, which can be used by the research community for further studies.
In this paper, we take stock of the current state of summarization datasets and explore how different factors of datasets influence the generalization behaviour of neural extractive summarization models. Specifically, we first propose several properties of datasets, which matter for the generalization of summarization models. Then we build the connection between priors residing in datasets and model designs, analyzing how different properties of datasets influence the choices of model structure design and training methods. Finally, by taking a typical dataset as an example, we rethink the process of the model design based on the experience of the above analysis. We demonstrate that when we have a deep understanding of the characteristics of datasets, a simple approach can bring significant improvements to the existing state-of-the-art model.
We construct Global Voices, a multilingual dataset for evaluating cross-lingual summarization methods. We extract social-network descriptions of Global Voices news articles to cheaply collect evaluation data for into-English and from-English summarization in 15 languages. Especially, for the into-English summarization task, we crowd-source a high-quality evaluation dataset based on guidelines that emphasize accuracy, coverage, and understandability. To ensure the quality of this dataset, we collect human ratings to filter out bad summaries, and conduct a survey on humans, which shows that the remaining summaries are preferred over the social-network summaries. We study the effect of translation quality in cross-lingual summarization, comparing a translate-then-summarize approach with several baselines. Our results highlight the limitations of the ROUGE metric that are overlooked in monolingual summarization.
Emerged as one of the best performing techniques for extractive summarization, determinantal point processes select a most probable set of summary sentences according to a probabilistic measure defined by respectively modeling sentence prominence and pairwise repulsion. Traditionally, both aspects are modelled using shallow and linguistically informed features, but the rise of deep contextualized representations raises an interesting question. Whether, and to what extent, could contextualized sentence representations be used to improve the DPP framework? Our findings suggest that, despite the success of deep semantic representations, it remains necessary to combine them with surface indicators for effective identification of summary-worthy sentences.
While recent work in abstractive summarization has resulted in higher scores in automatic metrics, there is little understanding on how these systems combine information taken from multiple document sentences. In this paper, we analyze the outputs of five state-of-the-art abstractive summarizers, focusing on summary sentences that are formed by sentence fusion. We ask assessors to judge the grammaticality, faithfulness, and method of fusion for summary sentences. Our analysis reveals that system sentences are mostly grammatical, but often fail to remain faithful to the original article.
Concept maps are visual summaries, structured as directed graphs: important concepts from a dataset are displayed as vertexes, and edges between vertexes show natural language descriptions of the relationships between the concepts on the map. Thus far, preliminary attempts at automatically creating concept maps have focused on building static summaries. However, in interactive settings, users will need to dynamically investigate particular relationships between pairs of concepts. For instance, a historian using a concept map browser might decide to investigate the relationship between two politicians in a news archive. We present a model which responds to such queries by returning one or more short, importance-ranked, natural language descriptions of the relationship between two requested concepts, for display in a visual interface. Our model is trained on a new public dataset, collected for this task.
Extractive summarization selects and concatenates the most essential text spans in a document. Most, if not all, neural approaches use sentences as the elementary unit to select content for summarization. However, semantic segments containing supplementary information or descriptive details are often nonessential in the generated summaries. In this work, we propose to exploit discourse-level segmentation as a finer-grained means to more precisely pinpoint the core content in a document. We investigate how the sub-sentential segmentation improves extractive summarization performance when content selection is modeled through two basic neural network architectures and a deep bi-directional transformer. Experiment results on the CNN/Daily Mail dataset show that discourse-level segmentation is effective in both cases. In particular, we achieve state-of-the-art performance when discourse-level segmentation is combined with our adapted contextual representation model.
We present the results of the Machine Reading for Question Answering (MRQA) 2019 shared task on evaluating the generalization capabilities of reading comprehension systems. In this task, we adapted and unified 18 distinct question answering datasets into the same format. Among them, six datasets were made available for training, six datasets were made available for development, and the rest were hidden for final evaluation. Ten teams submitted systems, which explored various ideas including data sampling, multi-task learning, adversarial training and ensembling. The best system achieved an average F1 score of 72.5 on the 12 held-out datasets, 10.7 absolute points higher than our initial baseline based on BERT.
Most machine reading comprehension (MRC) models separately handle encoding and matching with different network architectures. In contrast, pretrained language models with Transformer layers, such as GPT (Radford et al., 2018) and BERT (Devlin et al., 2018), have achieved competitive performance on MRC. A research question that naturally arises is: apart from the benefits of pre-training, how many performance gain comes from the unified network architecture. In this work, we evaluate and analyze unifying encoding and matching components with Transformer for the MRC task. Experimental results on SQuAD show that the unified model outperforms previous networks that separately treat encoding and matching. We also introduce a metric to inspect whether a Transformer layer tends to perform encoding or matching. The analysis results show that the unified model learns different modeling strategies compared with previous manually-designed models.
Machine reading comprehension is a task related to Question-Answering where questions are not generic in scope but are related to a particular document. Recently very large corpora (SQuAD, MS MARCO) containing triplets (document, question, answer) were made available to the scientific community to develop supervised methods based on deep neural networks with promising results. These methods need very large training corpus to be efficient, however such kind of data only exists for English and Chinese at the moment. The aim of this study is the development of such resources for other languages by proposing to generate in a semi-automatic way questions from the semantic Frame analysis of large corpora. The collect of natural questions is reduced to a validation/test set. We applied this method on the CALOR-Frame French corpus to develop the CALOR-QUEST resource presented in this paper.
We focus on multiple-choice question answering (QA) tasks in subject areas such as science, where we require both broad background knowledge and the facts from the given subject-area reference corpus. In this work, we explore simple yet effective methods for exploiting two sources of external knowledge for subject-area QA. The first enriches the original subject-area reference corpus with relevant text snippets extracted from an open-domain resource (i.e., Wikipedia) that cover potentially ambiguous concepts in the question and answer options. As in other QA research, the second method simply increases the amount of training data by appending additional in-domain subject-area instances. Experiments on three challenging multiple-choice science QA tasks (i.e., ARC-Easy, ARC-Challenge, and OpenBookQA) demonstrate the effectiveness of our methods: in comparison to the previous state-of-the-art, we obtain absolute gains in accuracy of up to 8.1%, 13.0%, and 12.8%, respectively. While we observe consistent gains when we introduce knowledge from Wikipedia, we find that employing additional QA training instances is not uniformly helpful: performance degrades when the added instances exhibit a higher level of difficulty than the original training data. As one of the first studies on exploiting unstructured external knowledge for subject-area QA, we hope our methods, observations, and discussion of the exposed limitations may shed light on further developments in the area.
In conversational machine comprehension, it has become one of the research hotspots integrating conversational history information through question reformulation for obtaining better answers. However, the existing question reformulation models are trained only using supervised question labels annotated by annotators without considering any feedback information from answers. In this paper, we propose a novel Answer-Supervised Question Reformulation (ASQR) model for enhancing conversational machine comprehension with reinforcement learning technology. ASQR utilizes a pointer-copy-based question reformulation model as an agent, takes an action to predict the next word, and observes a reward for the whole sentence state after generating the end-of-sequence token. The experimental results on QuAC dataset prove that our ASQR model is more effective in conversational machine comprehension. Moreover, pretraining is essential in reinforcement learning models, so we provide a high-quality annotated dataset for question reformulation by sampling a part of QuAC dataset.
A key challenge of multi-hop question answering (QA) in the open-domain setting is to accurately retrieve the supporting passages from a large corpus. Existing work on open-domain QA typically relies on off-the-shelf information retrieval (IR) techniques to retrieve answer passages, i.e., the passages containing the groundtruth answers. However, IR-based approaches are insufficient for multi-hop questions, as the topic of the second or further hops is not explicitly covered by the question. To resolve this issue, we introduce a new subproblem of open-domain multi-hop QA, which aims to recognize the bridge (i.e., the anchor that links to the answer passage) from the context of a set of start passages with a reading comprehension model. This model, the bridge reasoner, is trained with a weakly supervised signal and produces the candidate answer passages for the passage reader to extract the answer. On the full-wiki HotpotQA benchmark, we significantly improve the baseline method by 14 point F1. Without using any memory inefficient contextual embeddings, our result is also competitive with the state-of-the-art that applies BERT in multiple modules.
Despite the remarkable progress on Machine Reading Comprehension (MRC) with the help of open-source datasets, recent studies indicate that most of the current MRC systems unfortunately suffer from weak robustness against adversarial samples. To address this issue, we attempt to take sentence syntax as the leverage in the answer predicting process which previously only takes account of phrase-level semantics. Furthermore, to better utilize the sentence syntax and improve the robustness, we propose a Syntactic Leveraging Network, which is designed to deal with adversarial samples by exploiting the syntactic elements of a question. The experiment results indicate that our method is promising for improving the generalization and robustness of MRC models against the influence of adversarial samples, with performance well-maintained.
A key component of successfully reading a passage of text is the ability to apply knowledge gained from the passage to a new situation. In order to facilitate progress on this kind of reading, we present ROPES, a challenging benchmark for reading comprehension targeting Reasoning Over Paragraph Effects in Situations. We target expository language describing causes and effects (e.g., “animal pollinators increase efficiency of fertilization in flowers”), as they have clear implications for new situations. A system is presented a background passage containing at least one of these relations, a novel situation that uses this background, and questions that require reasoning about effects of the relationships in the background passage in the context of the situation. We collect background passages from science textbooks and Wikipedia that contain such phenomena, and ask crowd workers to author situations, questions, and answers, resulting in a 14,322 question dataset. We analyze the challenges of this task and evaluate the performance of state-of-the-art reading comprehension models. The best model performs only slightly better than randomly guessing an answer of the correct type, at 61.6% F1, well below the human performance of 89.0%.
Conversational question generation is a novel area of NLP research which has a range of potential applications. This paper is first to presents a framework for conversational question generation that is unaware of the corresponding answers. To properly generate a question coherent to the grounding text and the current conversation history, the proposed framework first locates the focus of a question in the text passage, and then identifies the question pattern that leads the sequential generation of the words in a question. The experiments using the CoQA dataset demonstrate that the quality of generated questions greatly improves if the question foci and the question patterns are correctly identified. In addition, it was shown that the question foci, even estimated with a reasonable accuracy, could contribute to the quality improvement. These results established that our research direction may be promising, but at the same time revealed that the identification of question patterns is a challenging issue, and it has to be largely refined to achieve a better quality in the end-to-end automatic question generation.
We demonstrate the viability of knowledge transfer between two related tasks: machine reading comprehension (MRC) and query-based text summarization. Using an MRC model trained on the SQuAD1.1 dataset as a core system component, we first build an extractive query-based summarizer. For better precision, this summarizer also compresses the output of the MRC model using a novel sentence compression technique. We further leverage pre-trained machine translation systems to abstract our extracted summaries. Our models achieve state-of-the-art results on the publicly available CNN/Daily Mail and Debatepedia datasets, and can serve as simple yet powerful baselines for future systems. We also hope that these results will encourage research on transfer learning from large MRC corpora to query-based summarization.
We present a system for answering questions based on the full text of books (BookQA), which first selects book passages given a question at hand, and then uses a memory network to reason and predict an answer. To improve generalization, we pretrain our memory network using artificial questions generated from book sentences. We experiment with the recently published NarrativeQA corpus, on the subset of Who questions, which expect book characters as answers. We experimentally show that BERT-based retrieval and pretraining improve over baseline results significantly. At the same time, we confirm that NarrativeQA is a highly challenging data set, and that there is need for novel research in order to achieve high-precision BookQA results. We analyze some of the bottlenecks of the current approach, and we argue that more research is needed on text representation, retrieval of relevant passages, and reasoning, including commonsense knowledge.
Conversational machine comprehension requires deep understanding of the dialogue flow, and the prior work proposed FlowQA to implicitly model the context representations in reasoning for better understanding. This paper proposes to explicitly model the information gain through the dialogue reasoning in order to allow the model to focus on more informative cues. The proposed model achieves the state-of-the-art performance in a conversational QA dataset QuAC and sequential instruction understanding dataset SCONE, which shows the effectiveness of the proposed mechanism and demonstrate its capability of generalization to different QA models and tasks.
General Question Answering (QA) systems over texts require the multi-hop reasoning capability, i.e. the ability to reason with information collected from multiple passages to derive the answer. In this paper we conduct a systematic analysis to assess such an ability of various existing models proposed for multi-hop QA tasks. Specifically, our analysis investigates that whether providing the full reasoning chain of multiple passages, instead of just one final passage where the answer appears, could improve the performance of the existing QA models. Surprisingly, when using the additional evidence passages, the improvements of all the existing multi-hop reading approaches are rather limited, with the highest error reduction of 5.8% on F1 (corresponding to 1.3% improvement) from the BERT model. To better understand whether the reasoning chains indeed could help find the correct answers, we further develop a co-matching-based method that leads to 13.1% error reduction with passage chains when applied to two of our base readers (including BERT). Our results demonstrate the existence of the potential improvement using explicit multi-hop reasoning and the necessity to develop models with better reasoning abilities.
To improve the accuracy of predicate-argument structure (PAS) analysis, large-scale training data and knowledge for PAS analysis are indispensable. We focus on a specific domain, specifically Japanese blogs on driving, and construct two wide-coverage datasets as a form of QA using crowdsourcing: a PAS-QA dataset and a reading comprehension QA (RC-QA) dataset. We train a machine comprehension (MC) model based on these datasets to perform PAS analysis. Our experiments show that a stepwise training method is the most effective, which pre-trains an MC model based on the RC-QA dataset to acquire domain knowledge and then fine-tunes based on the PAS-QA dataset.
Machine reading comprehension, the task of evaluating a machine’s ability to comprehend a passage of text, has seen a surge in popularity in recent years. There are many datasets that are targeted at reading comprehension, and many systems that perform as well as humans on some of these datasets. Despite all of this interest, there is no work that systematically defines what reading comprehension is. In this work, we justify a question answering approach to reading comprehension and describe the various kinds of questions one might use to more fully test a system’s comprehension of a passage, moving beyond questions that only probe local predicate-argument structures. The main pitfall of this approach is that questions can easily have surface cues or other biases that allow a model to shortcut the intended reasoning process. We discuss ways proposed in current literature to mitigate these shortcuts, and we conclude with recommendations for future dataset collection efforts.
Multi-hop question answering (QA) requires an information retrieval (IR) system that can find multiple supporting evidence needed to answer the question, making the retrieval process very challenging. This paper introduces an IR technique that uses information of entities present in the initially retrieved evidence to learn to ‘hop’ to other relevant evidence. In a setting, with more than 5 million Wikipedia paragraphs, our approach leads to significant boost in retrieval performance. The retrieved evidence also increased the performance of an existing QA model (without any training) on the benchmark by 10.59 F1.
As the complexity of question answering (QA) datasets evolve, moving away from restricted formats like span extraction and multiple-choice (MC) to free-form answer generation, it is imperative to understand how well current metrics perform in evaluating QA. This is especially important as existing metrics (BLEU, ROUGE, METEOR, and F1) are computed using n-gram similarity and have a number of well-known drawbacks. In this work, we study the suitability of existing metrics in QA. For generative QA, we show that while current metrics do well on existing datasets, converting multiple-choice datasets into free-response datasets is challenging for current metrics. We also look at span-based QA, where F1 is a reasonable metric. We show that F1 may not be suitable for all extractive QA tasks depending on the answer types. Our study suggests that while current metrics may be suitable for existing QA datasets, they limit the complexity of QA datasets that can be created. This is especially true in the context of free-form QA, where we would like our models to be able to generate more complex and abstractive answers, thus necessitating new metrics that go beyond n-gram based matching. As a step towards a better QA metric, we explore using BERTScore, a recently proposed metric for evaluating translation, for QA. We find that although it fails to provide stronger correlation with human judgements, future work focused on tailoring a BERT-based metric to QA evaluation may prove fruitful.
The field of question answering (QA) has seen rapid growth in new tasks and modeling approaches in recent years. Large scale datasets and focus on challenging linguistic phenomena have driven development in neural models, some of which have achieved parity with human performance in limited cases. However, an examination of state-of-the-art model output reveals that a gap remains in reasoning ability compared to a human, and performance tends to degrade when models are exposed to less-constrained tasks. We are interested in more clearly defining the strengths and limitations of leading models across diverse QA challenges, intending to help future researchers with identifying pathways to generalizable performance. We conduct extensive qualitative and quantitative analyses on the results of four models across four datasets and relate common errors to model capabilities. We also illustrate limitations in the datasets we examine and discuss a way forward for achieving generalizable models and datasets that broadly test QA capabilities.
Popular QA benchmarks like SQuAD have driven progress on the task of identifying answer spans within a specific passage, with models now surpassing human performance. However, retrieving relevant answers from a huge corpus of documents is still a challenging problem, and places different requirements on the model architecture. There is growing interest in developing scalable answer retrieval models trained end-to-end, bypassing the typical document retrieval step. In this paper, we introduce Retrieval Question-Answering (ReQA), a benchmark for evaluating large-scale sentence-level answer retrieval models. We establish baselines using both neural encoding models as well as classical information retrieval techniques. We release our evaluation code to encourage further work on this challenging task.
Reading comprehension is one of the crucial tasks for furthering research in natural language understanding. A lot of diverse reading comprehension datasets have recently been introduced to study various phenomena in natural language, ranging from simple paraphrase matching and entity typing to entity tracking and understanding the implications of the context. Given the availability of many such datasets, comprehensive and reliable evaluation is tedious and time-consuming for researchers working on this problem. We present an evaluation server, ORB, that reports performance on seven diverse reading comprehension datasets, encouraging and facilitating testing a single model’s capability in understanding a wide variety of reading phenomena. The evaluation server places no restrictions on how models are trained, so it is a suitable test bed for exploring training paradigms and representation learning for general reading facility. As more suitable datasets are released, they will be added to the evaluation server. We also collect and include synthetic augmentations for these datasets, testing how well models can handle out-of-domain questions.
In this study, we investigate the employment of the pre-trained BERT language model to tackle question generation tasks. We introduce three neural architectures built on top of BERT for question generation tasks. The first one is a straightforward BERT employment, which reveals the defects of directly using BERT for text generation. Accordingly, we propose another two models by restructuring our BERT employment into a sequential manner for taking information from previous decoded results. Our models are trained and evaluated on the recent question-answering dataset SQuAD. Experiment results show that our best model yields state-of-the-art performance which advances the BLEU 4 score of the existing best models from 16.85 to 22.17.
Question Generation (QG) is a Natural Language Processing (NLP) task that aids advances in Question Answering (QA) and conversational assistants. Existing models focus on generating a question based on a text and possibly the answer to the generated question. They need to determine the type of interrogative word to be generated while having to pay attention to the grammar and vocabulary of the question. In this work, we propose Interrogative-Word-Aware Question Generation (IWAQG), a pipelined system composed of two modules: an interrogative word classifier and a QG model. The first module predicts the interrogative word that is provided to the second module to create the question. Owing to an increased recall of deciding the interrogative words to be used for the generated questions, the proposed model achieves new state-of-the-art results on the task of QG in SQuAD, improving from 46.58 to 47.69 in BLEU-1, 17.55 to 18.53 in BLEU-4, 21.24 to 22.33 in METEOR, and from 44.53 to 46.94 in ROUGE-L.
Although advances in neural architectures for NLP problems as well as unsupervised pre-training have led to substantial improvements on question answering and natural language inference, understanding of and reasoning over long texts still poses a substantial challenge. Here, we consider the task of question answering from full narratives (e.g., books or movie scripts), or their summaries, tackling the NarrativeQA challenge (NQA; Kocisky et al. (2018)). We introduce a heuristic extractive version of the data set, which allows us to approach the more feasible problem of answer extraction (rather than generation). We train systems for passage retrieval as well as answer span prediction using this data set. We use pre-trained BERT embeddings for injecting prior knowledge into our system. We show that our setup leads to state of the art performance on summary-level QA. On QA from full narratives, our model outperforms previous models on the METEOR metric. We analyze the relative contributions of pre-trained embeddings and the extractive training paradigm, and provide a detailed error analysis.
This paper describes our model for the reading comprehension task of the MRQA shared task. We propose CLER, which stands for Cross-task Learning with Expert Representation for the generalization of reading and understanding. To generalize its capabilities, the proposed model is composed of three key ideas: multi-task learning, mixture of experts, and ensemble. In-domain datasets are used to train and validate our model, and other out-of-domain datasets are used to validate the generalization of our model’s performances. In a submission run result, the proposed model achieved an average F1 score of 66.1 % in the out-of-domain setting, which is a 4.3 percentage point improvement over the official BERT baseline model.
The model submitted works as follows. When supplied a question and a passage it makes use of the BERT embedding along with the hierarchical attention model which consists of 2 parts, the co-attention and the self-attention, to locate a continuous span of the passage that is the answer to the question.
Adapting models to new domain without finetuning is a challenging problem in deep learning. In this paper, we utilize an adversarial training framework for domain generalization in Question Answering (QA) task. Our model consists of a conventional QA model and a discriminator. The training is performed in the adversarial manner, where the two models constantly compete, so that QA model can learn domain-invariant features. We apply this approach in MRQA Shared Task 2019 and show better performance compared to the baseline model.
With a large number of datasets being released and new techniques being proposed, Question answering (QA) systems have witnessed great breakthroughs in reading comprehension (RC)tasks. However, most existing methods focus on improving in-domain performance, leaving open the research question of how these mod-els and techniques can generalize to out-of-domain and unseen RC tasks. To enhance the generalization ability, we propose a multi-task learning framework that learns the shared representation across different tasks. Our model is built on top of a large pre-trained language model, such as XLNet, and then fine-tuned on multiple RC datasets. Experimental results show the effectiveness of our methods, with an average Exact Match score of 56.59 and an average F1 score of 68.98, which significantly improves the BERT-Large baseline by8.39 and 7.22, respectively
In this paper, we introduce a simple system Baidu submitted for MRQA (Machine Reading for Question Answering) 2019 Shared Task that focused on generalization of machine reading comprehension (MRC) models. Our system is built on a framework of pretraining and fine-tuning, namely D-NET. The techniques of pre-trained language models and multi-task learning are explored to improve the generalization of MRC models and we conduct experiments to examine the effectiveness of these strategies. Our system is ranked at top 1 of all the participants in terms of averaged F1 score. Our codes and models will be released at PaddleNLP.
To produce a domain-agnostic question answering model for the Machine Reading Question Answering (MRQA) 2019 Shared Task, we investigate the relative benefits of large pre-trained language models, various data sampling strategies, as well as query and context paraphrases generated by back-translation. We find a simple negative sampling technique to be particularly effective, even though it is typically used for datasets that include unanswerable questions, such as SQuAD 2.0. When applied in conjunction with per-domain sampling, our XLNet (Yang et al., 2019)-based submission achieved the second best Exact Match and F1 in the MRQA leaderboard competition.
Crowdsourcing is frequently employed to quickly and inexpensively obtain valuable linguistic annotations but is rarely used for parsing, likely due to the perceived difficulty of the task and the limited training of the available workers. This paper presents what is, to the best of our knowledge, the first published use of Mechanical Turk (or similar platform) to crowdsource parse trees. We pay Turkers to construct unlabeled dependency trees for 500 English sentences using an interactive graphical dependency tree editor, collecting 10 annotations per sentence. Despite not requiring any training, several of the more prolific workers meet or exceed 90% attachment agreement with the Penn Treebank (PTB) portion of our data, and, furthermore, for 72% of these PTB sentences, at least one Turker produces a perfect parse. Thus, we find that, supported with a simple graphical interface, people with presumably no prior experience can achieve surprisingly high degrees of accuracy on this task. To facilitate research into aggregation techniques for complex crowdsourced annotations, we publicly release our annotated corpus.
This paper presents research on word familiarity rate estimation using the ‘Word List by Semantic Principles’. We collected rating information on 96,557 words in the ‘Word List by Semantic Principles’ via Yahoo! crowdsourcing. We asked 3,392 subject participants to use their introspection to rate the familiarity of words based on the five perspectives of ‘KNOW’, ‘WRITE’, ‘READ’, ‘SPEAK’, and ‘LISTEN’, and each word was rated by at least 16 subject participants. We used Bayesian linear mixed models to estimate the word familiarity rates. We also explored the ratings with the semantic labels used in the ‘Word List by Semantic Principles’.
Detecting event mentions is the first step in event extraction from text and annotating them is a notoriously difficult task. Evaluating annotator consistency is crucial when building datasets for mention detection. When event mentions are allowed to cover many tokens, annotators may disagree on their span, which means that overlapping annotations may then refer to the same event or to different events. This paper explores different fuzzy-matching functions which aim to resolve this ambiguity. The functions extract the sets of syntactic heads present in the annotations, use the Dice coefficient to measure the similarity between sets and return a judgment based on a given threshold. The functions are tested against the judgment of a human evaluator and a comparison is made between sets of tokens and sets of syntactic heads. The best-performing function is a head-based function that is found to agree with the human evaluator in 89% of cases.
The target outputs of many NLP tasks are word sequences. To collect the data for training and evaluating models, the crowd is a cheaper and easier to access than the oracle. To ensure the quality of the crowdsourced data, people can assign multiple workers to one question and then aggregate the multiple answers with diverse quality into a golden one. How to aggregate multiple crowdsourced word sequences with diverse quality is a curious and challenging problem. People need a dataset for addressing this problem. We thus create a dataset (CrowdWSA2019) which contains the translated sentences generated from multiple workers. We provide three approaches as the baselines on the task of extractive word sequence aggregation. Specially, one of them is an original one we propose which models the reliability of workers. We also discuss some issues on ground truth creation of word sequences which can be addressed based on this dataset.
Recent advancements in machine reading and listening comprehension involve the annotation of long texts. Such tasks are typically time consuming, making crowd-annotations an attractive solution, yet their complexity often makes such a solution unfeasible. In particular, a major concern is that crowd annotators may be tempted to skim through long texts, and answer questions without reading thoroughly. We present a case study of adapting this type of task to the crowd. The task is to identify claims in a several minute long debate speech. We show that sentence-by-sentence annotation does not scale and that labeling only a subset of sentences is insufficient. Instead, we propose a scheme for effectively performing the full, complex task with crowd annotators, allowing the collection of large scale annotated datasets. We believe that the encountered challenges and pitfalls, as well as lessons learned, are relevant in general when collecting data for large scale natural language understanding (NLU) tasks.
We propose a method of machine-assisted annotation for the identification of tension development, annotating whether the tension is increasing, decreasing, or staying unchanged. We use a neural network based prediction model, whose predicted results are given to the annotators as initial values for the options that they are asked to choose. By presenting such initial values to the annotators, the annotation task becomes an evaluation task where the annotators inspect whether or not the predicted results are correct. To demonstrate the effectiveness of our method, we performed the annotation task in both in-house and crowdsourced environments. For the crowdsourced environment, we compared the annotation results with and without our method of machine-assisted annotation. We find that the results with our method showed a higher agreement to the gold standard than those without, though our method had little effect at reducing the time for annotation. Our codes for the experiment are made publicly available.
Code-switching refers to the alternation of two or more languages in a conversation or utterance and is common in multilingual communities across the world. Building code-switched speech and natural language processing systems are challenging due to the lack of annotated speech and text data. We present a speech annotation interface CoSSAT, which helps annotators transcribe code-switched speech faster, more easily and more accurately than a traditional interface, by displaying candidate words from monolingual speech recognizers. We conduct a user study on the transcription of Hindi-English code-switched speech with 10 annotators and describe quantitative and qualitative results.
Pretrained deep contextual representations have advanced the state-of-the-art on various commonsense NLP tasks, but we lack a concrete understanding of the capability of these models. Thus, we investigate and challenge several aspects of BERT’s commonsense representation abilities. First, we probe BERT’s ability to classify various object attributes, demonstrating that BERT shows a strong ability in encoding various commonsense features in its embedding space, but is still deficient in many areas. Next, we show that, by augmenting BERT’s pretraining data with additional data related to the deficient attributes, we are able to improve performance on a downstream commonsense reasoning task while using a minimal amount of data. Finally, we develop a method of fine-tuning knowledge graphs embeddings alongside BERT and show the continued importance of explicit knowledge graphs.
This paper proposes a hybrid neural network(HNN) model for commonsense reasoning. An HNN consists of two component models, a masked language model and a semantic similarity model, which share a BERTbased contextual encoder but use different model-specific input and output layers. HNN obtains new state-of-the-art results on three classic commonsense reasoning tasks, pushing the WNLI benchmark to 89%, the Winograd Schema Challenge (WSC) benchmark to 75.1%, and the PDP60 benchmark to 90.0%. An ablation study shows that language models and semantic similarity models are complementary approaches to commonsense reasoning, and HNN effectively combines the strengths of both. The code and pre-trained models will be publicly available at https: //github.com/namisan/mt-dnn.
Non-extractive commonsense QA remains a challenging AI task, as it requires systems to reason about, synthesize, and gather disparate pieces of information, in order to generate responses to queries. Recent approaches on such tasks show increased performance, only when models are either pre-trained with additional information or when domain-specific heuristics are used, without any special consideration regarding the knowledge resource type. In this paper, we perform a survey of recent commonsense QA methods and we provide a systematic analysis of popular knowledge resources and knowledge-integration methods, across benchmarks from multiple commonsense datasets. Our results and analysis show that attention-based injection seems to be a preferable choice for knowledge integration and that the degree of domain overlap, between knowledge bases and datasets, plays a crucial role in determining model success.
Pretrained language models, such as BERT and RoBERTa, have shown large improvements in the commonsense reasoning benchmark COPA. However, recent work found that many improvements in benchmarks of natural language understanding are not due to models learning the task, but due to their increasing ability to exploit superficial cues, such as tokens that occur more often in the correct answer than the wrong one. Are BERT’s and RoBERTa’s good performance on COPA also caused by this? We find superficial cues in COPA, as well as evidence that BERT exploits these cues.To remedy this problem, we introduce Balanced COPA, an extension of COPA that does not suffer from easy-to-exploit single token cues. We analyze BERT’s and RoBERTa’s performance on original and Balanced COPA, finding that BERT relies on superficial cues when they are present, but still achieves comparable performance once they are made ineffective, suggesting that BERT learns the task to a certain degree when forced to. In contrast, RoBERTa does not appear to rely on superficial cues.
We consider the problem of extracting from text commonsense knowledge pertaining to human senses such as sound and smell. First, we consider the problem of recognizing mentions of human senses in text. Our contribution is a method for acquiring labeled data. Experiments show the effectiveness of our proposed data labeling approach when used with standard machine learning models on the task of sense recognition in text. Second, we propose to extract novel, common sense relationships pertaining to sense perception concepts. Our contribution is a process for generating labeled data by leveraging large corpora and crowdsourcing questionnaires.
Complex questions often require combining multiple facts to correctly answer, particularly when generating detailed explanations for why those answers are correct. Combining multiple facts to answer questions is often modeled as a “multi-hop” graph traversal problem, where a given solver must find a series of interconnected facts in a knowledge graph that, taken together, answer the question and explain the reasoning behind that answer. Multi-hop inference currently suffers from semantic drift, or the tendency for chains of reasoning to “drift”’ to unrelated topics, and this semantic drift greatly limits the number of facts that can be combined in both free text or knowledge base inference. In this work we present our effort to mitigate semantic drift by extracting large high-confidence multi-hop inference patterns, generated by abstracting large-scale explanatory structure from a corpus of detailed explanations. We represent these inference patterns as sets of generalized constraints over sentences represented as rows in a knowledge base of semi-structured tables. We present a prototype tool for identifying common inference patterns from corpora of semi-structured explanations, and use it to successfully extract 67 inference patterns from a “matter” subset of standardized elementary science exam questions that span scientific and world knowledge.
This paper reports on the results of the shared tasks of the COIN workshop at EMNLP-IJCNLP 2019. The tasks consisted of two machine comprehension evaluations, each of which tested a system’s ability to answer questions/queries about a text. Both evaluations were designed such that systems need to exploit commonsense knowledge, for example, in the form of inferences over information that is available in the common ground but not necessarily mentioned in the text. A total of five participating teams submitted systems for the shared tasks, with the best submitted system achieving 90.6% accuracy and 83.7% F1-score on task 1 and task 2, respectively.
This paper describes our model for COmmonsense INference in Natural Language Processing (COIN) shared task 1: Commonsense Inference in Everyday Narrations. This paper explores the use of Bidirectional Encoder Representations from Transformers(BERT) along with external relational knowledge from ConceptNet to tackle the problem of commonsense inference. The input passage, question, and answer are augmented with relational knowledge from ConceptNet. Using this technique we are able to achieve an accuracy of 73.3 % on the official test data.
In this paper, we describe our system for COIN 2019 Shared Task 1: Commonsense Inference in Everyday Narrations. We show the power of leveraging state-of-the-art pre-trained language models such as BERT(Bidirectional Encoder Representations from Transformers) and XLNet over other Commonsense Knowledge Base Resources such as ConceptNet and NELL for modeling machine comprehension. We used an ensemble of BERT-Large and XLNet-Large. Experimental results show that our model give substantial improvements over the baseline and other systems incorporating knowledge bases. We bagged 2nd position on the final test set leaderboard with an accuracy of 90.5%
We introduce a simple yet effective method of integrating contextual embeddings with commonsense graph embeddings, dubbed BERT Infused Graphs: Matching Over Other embeDdings. First, we introduce a preprocessing method to improve the speed of querying knowledge bases. Then, we develop a method of creating knowledge embeddings from each knowledge base. We introduce a method of aligning tokens between two misaligned tokenization methods. Finally, we contribute a method of contextualizing BERT after combining with knowledge base embeddings. We also show BERTs tendency to correct lower accuracy question types. Our model achieves a higher accuracy than BERT, and we score fifth on the official leaderboard of the shared task and score the highest without any additional language model pretraining.
To solve the shared tasks of COIN: COmmonsense INference in Natural Language Processing) Workshop in , we need explore the impact of knowledge representation in modeling commonsense knowledge to boost performance of machine reading comprehension beyond simple text matching. There are two approaches to represent knowledge in the low-dimensional space. The first is to leverage large-scale unsupervised text corpus to train fixed or contextual language representations. The second approach is to explicitly express knowledge into a knowledge graph (KG), and then fit a model to represent the facts in the KG. We have experimented both (a) improving the fine-tuning of pre-trained language models on a task with a small dataset size, by leveraging datasets of similar tasks; and (b) incorporating the distributional representations of a KG onto the representations of pre-trained language models, via simply concatenation or multi-head attention. We find out that: (a) for task 1, first fine-tuning on larger datasets like RACE (Lai et al., 2017) and SWAG (Zellersetal.,2018), and then fine-tuning on the target task improve the performance significantly; (b) for task 2, we find out the incorporating a KG of commonsense knowledge, WordNet (Miller, 1995) into the Bert model (Devlin et al., 2018) is helpful, however, it will hurts the performace of XLNET (Yangetal.,2019), a more powerful pre-trained model. Our approaches achieve the state-of-the-art results on both shared task’s official test data, outperforming all the other submissions.
This paper describes our system for COIN Shared Task 1: Commonsense Inference in Everyday Narrations. To inject more external knowledge to better reason over the narrative passage, question and answer, the system adopts a stagewise fine-tuning method based on pre-trained BERT model. More specifically, the first stage is to fine-tune on addi- tional machine reading comprehension dataset to learn more commonsense knowledge. The second stage is to fine-tune on target-task (MCScript2.0) with MCScript (2018) dataset assisted. Experimental results show that our system achieves significant improvements over the baseline systems with 84.2% accuracy on the official test dataset.
Natural language communication between machines and humans are still constrained. The article addresses a gap in natural language understanding about actions, specifically that of understanding commands. We propose a new method for commonsense inference (grounding) of high-level natural language commands into specific action commands for further execution by a robotic system. The method allows to build a knowledge base that consists of a large set of commonsense inferences. The preliminary results have been presented.
Typical event sequences are an important class of commonsense knowledge. Formalizing the task as the generation of a next event conditioned on a current event, previous work in event prediction employs sequence-to-sequence (seq2seq) models. However, what can happen after a given event is usually diverse, a fact that can hardly be captured by deterministic models. In this paper, we propose to incorporate a conditional variational autoencoder (CVAE) into seq2seq for its ability to represent diverse next events as a probabilistic distribution. We further extend the CVAE-based seq2seq with a reconstruction mechanism to prevent the model from concentrating on highly typical events. To facilitate fair and systematic evaluation of the diversity-aware models, we also extend existing evaluation datasets by tying each current event to multiple next events. Experiments show that the CVAE-based models drastically outperform deterministic models in terms of precision and that the reconstruction mechanism improves the recall of CVAE-based models without sacrificing precision.
Modeling semantic plausibility requires commonsense knowledge about the world and has been used as a testbed for exploring various knowledge representations. Previous work has focused specifically on modeling physical plausibility and shown that distributional methods fail when tested in a supervised setting. At the same time, distributional models, namely large pretrained language models, have led to improved results for many natural language understanding tasks. In this work, we show that these pretrained language models are in fact effective at modeling physical plausibility in the supervised setting. We therefore present the more difficult problem of learning to model physical plausibility directly from text. We create a training set by extracting attested events from a large corpus, and we provide a baseline for training on these attested events in a self-supervised manner and testing on a physical plausibility task. We believe results could be further improved by injecting explicit commonsense knowledge into a distributional model.
Understanding common sense is important for effective natural language reasoning. One type of common sense is how two objects compare on physical properties such as size and weight: e.g., ‘is a house bigger than a person?’. We probe whether pre-trained representations capture comparisons and find they, in fact, have higher accuracy than previous approaches. They also generalize to comparisons involving objects not seen during training. We investigate how such comparisons are made: models learn a consistent ordering over all the objects in the comparisons. Probing models have significantly higher accuracy than those baseline models which use dataset artifacts: e.g., memorizing some words are larger than any other word.
New conversation topics and functionalities are constantly being added to conversational AI agents like Amazon Alexa and Apple Siri. As data collection and annotation is not scalable and is often costly, only a handful of examples for the new functionalities are available, which results in poor generalization performance. We formulate it as a Few-Shot Integration (FSI) problem where a few examples are used to introduce a new intent. In this paper, we study six feature space data augmentation methods to improve classification performance in FSI setting in combination with both supervised and unsupervised representation learning methods such as BERT. Through realistic experiments on two public conversational datasets, SNIPS, and the Facebook Dialog corpus, we show that data augmentation in feature space provides an effective way to improve intent classification performance in few-shot setting beyond traditional transfer learning approaches. In particular, we show that (a) upsampling in latent space is a competitive baseline for feature space augmentation (b) adding the difference between two examples to a new example is a simple yet effective data augmentation method.
To overcome the lack of annotated resources in less-resourced languages, recent approaches have been proposed to perform unsupervised language adaptation. In this paper, we explore three recent proposals: Adversarial Training, Sentence Encoder Alignment and Shared-Private Architecture. We highlight the differences of these approaches in terms of unlabeled data requirements and capability to overcome additional domain shift in the data. A comparative analysis in two different tasks is conducted, namely on Sentiment Classification and Natural Language Inference. We show that adversarial training methods are more suitable when the source and target language datasets contain other variations in content besides the language shift. Otherwise, sentence encoder alignment methods are very effective and can yield scores on the target language that are close to the source language scores.
At present, different deep learning models are presenting high accuracy on popular inference datasets such as SNLI, MNLI, and SciTail. However, there are different indicators that those datasets can be exploited by using some simple linguistic patterns. This fact poses difficulties to our understanding of the actual capacity of machine learning models to solve the complex task of textual inference. We propose a new set of syntactic tasks focused on contradiction detection that require specific capacities over linguistic logical forms such as: Boolean coordination, quantifiers, definite description, and counting operators. We evaluate two kinds of deep learning models that implicitly exploit language structure: recurrent models and the Transformer network BERT. We show that although BERT is clearly more efficient to generalize over most logical forms, there is space for improvement when dealing with counting operators. Since the syntactic tasks can be implemented in different languages, we show a successful case of cross-lingual transfer learning between English and Portuguese.
Word embeddings are an essential component in a wide range of natural language processing applications. However, distributional semantic models are known to struggle when only a small number of context sentences are available. Several methods have been proposed to obtain higher-quality vectors for these words, leveraging both this context information and sometimes the word forms themselves through a hybrid approach. We show that the current tasks do not suffice to evaluate models that use word-form information, as such models can easily leverage word forms in the training data that are related to word forms in the test data. We introduce 3 new tasks, allowing for a more balanced comparison between models. Furthermore, we show that hyperparameters that have largely been ignored in previous work can consistently improve the performance of both baseline and advanced models, achieving a new state of the art on 4 out of 6 tasks.
Many architectures for multi-task learning (MTL) have been proposed to take advantage of transfer among tasks, often involving complex models and training procedures. In this paper, we ask if the sentence-level representations learned in previous approaches provide significant benefit beyond that provided by simply improving word-based representations. To investigate this question, we consider three techniques that ignore sequence information: a syntactically-oblivious pooling encoder, pre-trained non-contextual word embeddings, and unigram generative regularization. Compared to a state-of-the-art MTL approach to textual inference, the simple techniques we use yield similar performance on a universe of task combinations while reducing training time and model size.
Multilingual transfer learning can benefit both high- and low-resource languages, but the source of these improvements is not well understood. Cananical Correlation Analysis (CCA) of the internal representations of a pre- trained, multilingual BERT model reveals that the model partitions representations for each language rather than using a common, shared, interlingual space. This effect is magnified at deeper layers, suggesting that the model does not progressively abstract semantic con- tent while disregarding languages. Hierarchical clustering based on the CCA similarity scores between languages reveals a tree structure that mirrors the phylogenetic trees hand- designed by linguists. The subword tokenization employed by BERT provides a stronger bias towards such structure than character- and word-level tokenizations. We release a subset of the XNLI dataset translated into an additional 14 languages at https://www.github.com/salesforce/xnli_extension to assist further research into multilingual representations.
Entities, which refer to distinct objects in the real world, can be viewed as language universals and used as effective signals to generate less ambiguous semantic representations and align multiple languages. We propose a novel method, CLEW, to generate cross-lingual data that is a mix of entities and contextual words based on Wikipedia. We replace each anchor link in the source language with its corresponding entity title in the target language if it exists, or in the source language otherwise. A cross-lingual joint entity and word embedding learned from this kind of data not only can disambiguate linkable entities but can also effectively represent unlinkable entities. Because this multilingual common space directly relates the semantics of contextual words in the source language to that of entities in the target language, we leverage it for unsupervised cross-lingual entity linking. Experimental results show that CLEW significantly advances the state-of-the-art: up to 3.1% absolute F-score gain for unsupervised cross-lingual entity linking. Moreover, it provides reliable alignment on both the word/entity level and the sentence level, and thus we use it to mine parallel sentences for all (302, 2) language pairs in Wikipedia.
We present a novel framework to deal with relation extraction tasks in cases where there is complete lack of supervision, either in the form of gold annotations, or relations from a knowledge base. Our approach leverages syntactic parsing and pre-trained word embeddings to extract few but precise relations, which are then used to annotate a larger corpus, in a manner identical to distant supervision. The resulting data set is employed to fine tune a pre-trained BERT model in order to perform relation extraction. Empirical evaluation on four data sets from the biomedical domain shows that our method significantly outperforms two simple baselines for unsupervised relation extraction and, even if not using any supervision at all, achieves slightly worse results than the state-of-the-art in three out of four data sets. Importantly, we show that it is possible to successfully fine tune a large pretrained language model with noisy data, as opposed to previous works that rely on gold data for fine tuning.
The performance of deep neural models can deteriorate substantially when there is a domain shift between training and test data. For example, the pre-trained BERT model can be easily fine-tuned with just one additional output layer to create a state-of-the-art model for a wide range of tasks. However, the fine-tuned BERT model suffers considerably at zero-shot when applied to a different domain. In this paper, we present a novel two-step domain adaptation framework based on curriculum learning and domain-discriminative data selection. The domain adaptation is conducted in a mostly unsupervised manner using a small target domain validation set for hyper-parameter tuning. We tested the framework on four large public datasets with different domain similarities and task types. Our framework outperforms a popular discrepancy-based domain adaptation method on most transfer tasks while consuming only a fraction of the training budget.
Active learning (AL) for machine translation (MT) has been well-studied for the phrase-based MT paradigm. Several AL algorithms for data sampling have been proposed over the years. However, given the rapid advancement in neural methods, these algorithms have not been thoroughly investigated in the context of neural MT (NMT). In this work, we address this missing aspect by conducting a systematic comparison of different AL methods in a simulated AL framework. Our experimental setup to compare different AL methods uses: i) State-of-the-art NMT architecture to achieve realistic results; and ii) the same dataset (WMT’13 English-Spanish) to have fair comparison across different methods. We then demonstrate how recent advancements in unsupervised pre-training and paraphrastic embedding can be used to improve existing AL methods. Finally, we propose a neural extension for an AL sampling method used in the context of phrase-based MT - Round Trip Translation Likelihood (RTTL). RTTL uses a bidirectional translation model to estimate the loss of information during translation and outperforms previous methods.
Semantic parsers are used to convert user’s natural language commands to executable logical form in intelligent personal agents. Labeled datasets required to train such parsers are expensive to collect, and are never comprehensive. As a result, for effective post-deployment domain adaptation and personalization, semantic parsers are continuously retrained to learn new user vocabulary and paraphrase variety. However, state-of-the art attention based neural parsers are slow to retrain which inhibits real time domain adaptation. Secondly, these parsers do not leverage numerous paraphrases already present in the training dataset. Designing parsers which can simultaneously maintain high accuracy and fast retraining time is challenging. In this paper, we present novel paraphrase attention based sequence-to-sequence/tree parsers which support fast near real time retraining. In addition, our parsers often boost accuracy by jointly modeling the semantic dependencies of paraphrases. We evaluate our model on benchmark datasets to demonstrate upto 9X speedup in retraining time compared to existing parsers, as well as achieving state-of-the-art accuracy.
Historical text normalization often relies on small training datasets. Recent work has shown that multi-task learning can lead to significant improvements by exploiting synergies with related datasets, but there has been no systematic study of different multi-task learning architectures. This paper evaluates 63 multi-task learning configurations for sequence-to-sequence-based historical text normalization across ten datasets from eight languages, using autoencoding, grapheme-to-phoneme mapping, and lemmatization as auxiliary tasks. We observe consistent, significant improvements across languages when training data for the target task is limited, but minimal or no improvements when training data is abundant. We also show that zero-shot learning outperforms the simple, but relatively strong, identity baseline.
Recent research on cross-lingual transfer show state-of-the-art results on benchmark datasets using pre-trained language representation models (PLRM) like BERT. These results are achieved with the traditional training approaches, such as Zero-shot with no data, Translate-train or Translate-test with machine translated data. In this work, we propose an approach of “Multilingual Co-training” (MCT) where we augment the expert annotated dataset in the source language (English) with the corresponding machine translations in the target languages (e.g. Arabic, Spanish) and fine-tune the PLRM jointly. We observe that the proposed approach provides consistent gains in the performance of BERT for multiple benchmark datasets (e.g. 1.0% gain on MLDocs, and 1.2% gain on XNLI over translate-train with BERT), while requiring a single model for multiple languages. We further consider a FAQ dataset where the available English test dataset is translated by experts into Arabic and Spanish. On such a dataset, we observe an average gain of 4.9% over all other cross-lingual transfer protocols with BERT. We further observe that domain-specific joint pre-training of the PLRM using HR policy documents in English along with the machine translations in the target languages, followed by the joint finetuning, provides a further improvement of 2.8% in average accuracy.
Over the past year, the emergence of transfer learning with large-scale language models (LM) has led to dramatic performance improvements across a broad range of natural language understanding tasks. However, the size and memory footprint of these large LMs often makes them difficult to deploy in many scenarios (e.g. on mobile phones). Recent research points to knowledge distillation as a potential solution, showing that when training data for a given task is abundant, it is possible to distill a large (teacher) LM into a small task-specific (student) network with minimal loss of performance. However, when such data is scarce, there remains a significant performance gap between large pretrained LMs and smaller task-specific models, even when training via distillation. In this paper, we bridge this gap with a novel training approach, called generation-distillation, that leverages large finetuned LMs in two ways: (1) to generate new (unlabeled) training examples, and (2) to distill their knowledge into a small network using these examples. Across three low-resource text classification datsets, we achieve comparable performance to BERT while using 300 times fewer parameters, and we outperform prior approaches to distillation for text classification while using 3 times fewer parameters.
Statistical natural language inference (NLI) models are susceptible to learning dataset bias: superficial cues that happen to associate with the label on a particular dataset, but are not useful in general, e.g., negation words indicate contradiction. As exposed by several recent challenge datasets, these models perform poorly when such association is absent, e.g., predicting that “I love dogs.” contradicts “I don’t love cats.”. Our goal is to design learning algorithms that guard against known dataset bias. We formalize the concept of dataset bias under the framework of distribution shift and present a simple debiasing algorithm based on residual fitting, which we call DRiFt. We first learn a biased model that only uses features that are known to relate to dataset bias. Then, we train a debiased model that fits to the residual of the biased model, focusing on examples that cannot be predicted well by biased features only. We use DRiFt to train three high-performing NLI models on two benchmark datasets, SNLI and MNLI. Our debiased models achieve significant gains over baseline models on two challenge test sets, while maintaining reasonable performance on the original test sets.
Traditional text classifiers are limited to predicting over a fixed set of labels. However, in many real-world applications the label set is frequently changing. For example, in intent classification, new intents may be added over time while others are removed. We propose to address the problem of dynamic text classification by replacing the traditional, fixed-size output layer with a learned, semantically meaningful metric space. Here the distances between textual inputs are optimized to perform nearest-neighbor classification across overlapping label sets. Changing the label set does not involve removing parameters, but rather simply adding or removing support points in the metric space. Then the learned metric can be fine-tuned with only a few additional training examples. We demonstrate that this simple strategy is robust to changes in the label space. Furthermore, our results show that learning a non-Euclidean metric can improve performance in the low data regime, suggesting that further work on metric spaces may benefit low-resource research.
The Lottery Ticket Hypothesis suggests large, over-parameterized neural networks consist of small, sparse subnetworks that can be trained in isolation to reach a similar (or better) test accuracy. However, the initialization and generalizability of the obtained sparse subnetworks have been recently called into question. Our work focuses on evaluating the initialization of sparse subnetworks under distributional shifts. Specifically, we investigate the extent to which a sparse subnetwork obtained in a source domain can be re-trained in isolation in a dissimilar, target domain. In addition, we examine the effects of different initialization strategies at transfer-time. Our experiments show that sparse subnetworks obtained through lottery ticket training do not simply overfit to particular domains, but rather reflect an inductive bias of deep neural networks that can be exploited in multiple domains.
Cross-lingual dependency parsing involves transferring syntactic knowledge from one language to another. It is a crucial component for inducing dependency parsers in low-resource scenarios where no training data for a language exists. Using Faroese as the target language, we compare two approaches using annotation projection: first, projecting from multiple monolingual source models; second, projecting from a single polyglot model which is trained on the combination of all source languages. Furthermore, we reproduce multi-source projection (Tyers et al., 2018), in which dependency trees of multiple sources are combined. Finally, we apply multi-treebank modelling to the projected treebanks, in addition to or alternatively to polyglot modelling on the source side. We find that polyglot training on the source languages produces an overall trend of better results on the target language but the single best result for the target language is obtained by projecting from monolingual source parsing models and then training multi-treebank POS tagging and parsing models on the target side.
Short Answer Grading (SAG) is a task of scoring students’ answers in examinations. Most existing SAG systems predict scores based only on the answers, including the model used as base line in this paper, which gives the-state-of-the-art performance. But they ignore important evaluation criteria such as rubrics, which play a crucial role for evaluating answers in real-world situations. In this paper, we present a method to inject information from rubrics into SAG systems. We implement our approach on top of word-level attention mechanism to introduce the rubric information, in order to locate information in each answer that are highly related to the score. Our experimental results demonstrate that injecting rubric information effectively contributes to the performance improvement and that our proposed model outperforms the state-of-the-art SAG model on the widely used ASAP-SAS dataset under low-resource settings.
Supervised learning models are typically trained on a single dataset and the performance of these models rely heavily on the size of the dataset i.e., the amount of data available with ground truth. Learning algorithms try to generalize solely based on the data that it is presented with during the training. In this work, we propose an inductive transfer learning method that can augment learning models by infusing similar instances from different learning tasks in Natural Language Processing (NLP) domain. We propose to use instance representations from a source dataset, without inheriting anything else from the source learning model. Representations of the instances of source and target datasets are learned, retrieval of relevant source instances is performed using soft-attention mechanism and locality sensitive hashing and then augmented into the model during training on the target dataset. Therefore, while learning from a training data, we also simultaneously exploit and infuse relevant local instance-level information from an external data. Using this approach we have shown significant improvements over the baseline for three major news classification datasets. Experimental evaluations also show that the proposed approach reduces dependency on labeled data by a significant margin for comparable performance. With our proposed cross dataset learning procedure we show that one can achieve competitive/better performance than learning from a single dataset.
Grapheme-to-phoneme conversion (g2p) is the task of predicting the pronunciation of words from their orthographic representation. His- torically, g2p systems were transition- or rule- based, making generalization beyond a mono- lingual (high resource) domain impractical. Recently, neural architectures have enabled multilingual systems to generalize widely; however, all systems to date have been trained only on spelling-pronunciation pairs. We hy- pothesize that the sequences of IPA characters used to represent pronunciation do not capture its full nuance, especially when cleaned to fa- cilitate machine learning. We leverage audio data as an auxiliary modality in a multi-task training process to learn a more optimal inter- mediate representation of source graphemes; this is the first multimodal model proposed for multilingual g2p. Our approach is highly ef- fective: on our in-domain test set, our mul- timodal model reduces phoneme error rate to 2.46%, a more than 65% decrease compared to our implementation of a unimodal spelling- pronunciation model—which itself achieves state-of-the-art results on the Wiktionary test set. The advantages of the multimodal model generalize to wholly unseen languages, reduc- ing phoneme error rate on our out-of-domain test set to 6.39% from the unimodal 8.21%, a more than 20% relative decrease. Further- more, our training and test sets are composed primarily of low-resource languages, demon- strating that our multimodal approach remains useful when training data are constrained.
Knowledge distillation can effectively transfer knowledge from BERT, a deep language representation model, to traditional, shallow word embedding-based neural networks, helping them approach or exceed the quality of other heavyweight language representation models. As shown in previous work, critical to this distillation procedure is the construction of an unlabeled transfer dataset, which enables effective knowledge transfer. To create transfer set examples, we propose to sample from pretrained language models fine-tuned on task-specific text. Unlike previous techniques, this directly captures the purpose of the transfer set. We hypothesize that this principled, general approach outperforms rule-based techniques. On four datasets in sentiment classification, sentence similarity, and linguistic acceptability, we show that our approach improves upon previous methods. We outperform OpenAI GPT, a deep pretrained transformer, on three of the datasets, while using a single-layer bidirectional LSTM that runs at least ten times faster.
Recently, neural network models which automatically infer syntactic structure from raw text have started to achieve promising results. However, earlier work on unsupervised parsing shows large performance differences between non-neural models trained on corpora in different languages, even for comparable amounts of data. With that in mind, we train instances of the PRPN architecture (Shen et al., 2018)—one of these unsupervised neural network parsers—for Arabic, Chinese, English, and German. We find that (i) the model strongly outperforms trivial baselines and, thus, acquires at least some parsing ability for all languages; (ii) good hyperparameter values seem to be universal; (iii) how the model benefits from larger training set sizes depends on the corpus, with the model achieving the largest performance gains when increasing the number of sentences from 2,500 to 12,500 for English. In addition, we show that, by sharing parameters between the related languages German and English, we can improve the model’s unsupervised parsing F1 score by up to 4% in the low-resource setting.
Argument component extraction is a challenging and complex high-level semantic extraction task. As such, it is both expensive to annotate (meaning training data is limited and low-resource by nature), and hard for current-generation deep learning methods to model. In this paper, we reevaluate the performance of state-of-the-art approaches in both single- and multi-task learning settings using combinations of character-level, GloVe, ELMo, and BERT encodings using standard BiLSTM-CRF encoders. We use evaluation metrics that are more consistent with evaluation practice in named entity recognition to understand how well current baselines address this challenge and compare their performance to lower-level semantic tasks such as CoNLL named entity recognition. We find that performance utilizing various pre-trained representations and training methodologies often leaves a lot to be desired as it currently stands, and suggest future pathways for improvement.
Existing named entity recognition (NER) systems rely on large amounts of human-labeled data for supervision. However, obtaining large-scale annotated data is challenging particularly in specific domains like health-care, e-commerce and so on. Given the availability of domain specific knowledge resources, (e.g., ontologies, dictionaries), distant supervision is a solution to generate automatically labeled training data to reduce human effort. The outcome of distant supervision for NER, however, is often noisy. False positive and false negative instances are the main issues that reduce performance on this kind of auto-generated data. In this paper, we explore distant supervision in a supervised setup. We adopt a technique of partial annotation to address false negative cases and implement a reinforcement learning strategy with a neural network policy to identify false positive instances. Our results establish a new state-of-the-art on four benchmark datasets taken from different domains and different languages. We then go on to show that our model reduces the amount of manually annotated data required to perform NER in a new domain.
In this paper, a dialogue system for Hospital domain in Telugu, which is a resource-poor Dravidian language, has been built. It handles various hospital and doctor related queries. The main aim of this paper is to present an approach for modelling a dialogue system in a resource-poor language by combining linguistic and domain knowledge. Focusing on the question answering aspect of the dialogue system, we identified Question Classification and Query Processing as the two most important parts of the dialogue system. Our method combines deep learning techniques for question classification and computational rule-based analysis for query processing. Human evaluation of the system has been performed as there is no automated evaluation tool for dialogue systems in Telugu. Our system achieves a high overall rating along with a significantly accurate context-capturing method as shown in the results.
Cross-lingual entity linking (XEL) grounds named entities in a source language to an English Knowledge Base (KB), such as Wikipedia. XEL is challenging for most languages because of limited availability of requisite resources. However, many works on XEL have been on simulated settings that actually use significant resources (e.g. source language Wikipedia, bilingual entity maps, multilingual embeddings) that are not available in truly low-resource languages. In this work, we first examine the effect of these resource assumptions and quantify how much the availability of these resource affects overall quality of existing XEL systems. We next propose three improvements to both entity candidate generation and disambiguation that make better use of the limited resources we do have in resource-scarce scenarios. With experiments on four extremely low-resource languages, we show that our model results in gains of 6-20% end-to-end linking accuracy.
Multi-task learning and self-training are two common ways to improve a machine learning model’s performance in settings with limited training data. Drawing heavily on ideas from those two approaches, we suggest transductive auxiliary task self-training: training a multi-task model on (i) a combination of main and auxiliary task training data, and (ii) test instances with auxiliary task labels which a single-task version of the model has previously generated. We perform extensive experiments on 86 combinations of languages and tasks. Our results are that, on average, transductive auxiliary task self-training improves absolute accuracy by up to 9.56% over the pure multi-task model for dependency relation tagging and by up to 13.03% for semantic tagging.
We propose a weakly supervised neural model for Ad-hoc Cross-lingual Information Retrieval (CLIR) from low-resource languages. Low resource languages often lack relevance annotations for CLIR, and when available the training data usually has limited coverage for possible queries. In this paper, we design a model which does not require relevance annotations, instead it is trained on samples extracted from translation corpora as weak supervision. This model relies on an attention mechanism to learn spans in the foreign sentence that are relevant to the query. We report experiments on two low resource languages: Swahili and Tagalog, trained on less that 100k parallel sentences each. The proposed model achieves 19 MAP points improvement compared to using CNNs for feature extraction, 12 points improvement from machine translation-based CLIR, and up to 6 points improvement compared to probabilistic CLIR models.
Although the vast majority of knowledge bases (KBs) are heavily biased towards English, Wikipedias do cover very different topics in different languages. Exploiting this, we introduce a new multilingual dataset (X-WikiRE), framing relation extraction as a multilingual machine reading problem. We show that by leveraging this resource it is possible to robustly transfer models cross-lingually and that multilingual support significantly improves (zero-shot) relation extraction, enabling the population of low-resourced KBs from their well-populated counterparts.
In this paper we address a challenging cross-lingual name retrieval task. Given an English named entity query, we aim to find all name mentions in documents in low-resource languages. We present a novel method which relies on zero annotation or resources from the target language. By leveraging freely available, cross-lingual resources and a small amount of training data from another language, we are able to perform name retrieval on a new language without any additional training data. Our method proceeds in a multi-step process: first, we pre-train a language-independent orthographic encoder using Wikipedia inter-lingual links from dozens of languages. Next, we gather user expectations about important entities in an English comparable document and compare those expected entities with actual spans of the target language text in order to perform name finding. Our method shows 11.6% absolute F-score improvement over state-of-the-art methods.
We investigate whether off-the-shelf deep bidirectional sentence representations (Devlin et al., 2019) trained on a massively multilingual corpus (multilingual BERT) enable the development of an unsupervised universal dependency parser. This approach only leverages a mix of monolingual corpora in many languages and does not require any translation data making it applicable to low-resource languages. In our experiments we outperform the best CoNLL 2018 language-specific systems in all of the shared task’s six truly low-resource languages while using a single system. However, we also find that (i) parsing accuracy still varies dramatically when changing the training languages and (ii) in some target languages zero-shot transfer fails under all tested conditions, raising concerns on the ‘universality’ of the whole approach.
We report results from the SR’19 Shared Task, the second edition of a multilingual surface realisation task organised as part of the EMNLP’19 Workshop on Multilingual Surface Realisation. As in SR’18, the shared task comprised two tracks with different levels of complexity: (a) a shallow track where the inputs were full UD structures with word order information removed and tokens lemmatised; and (b) a deep track where additionally, functional words and morphological information were removed. The shallow track was offered in eleven, and the deep track in three languages. Systems were evaluated (a) automatically, using a range of intrinsic metrics, and (b) by human judges in terms of readability and meaning similarity. This report presents the evaluation results, along with descriptions of the SR’19 tracks, data and evaluation methods. For full descriptions of the participating systems, please see the separate system reports elsewhere in this volume.
Recent advances in deep learning have shown promises in solving complex combinatorial optimization problems, such as sorting variable-sized sequences. In this work, we take a step further and tackle the problem of ordering the elements of sequences that come with graph structures. Our solution adopts an encoder-decoder framework, in which the encoder is a graph neural network that learns the representation for each element, and the decoder predicts the ordering of each local neighborhood of the graph in turn. We apply our framework to multilingual surface realization, which is the task of ordering and completing sentences with their dependency parses given but without the ordering of words. Experiments show that our approach is much better for this task than prior works that do not consider graph structures. We participated in 2019 Surface Realization Shared Task (SR’19), and we ranked second out of 14 teams while outperforming those teams below by a large margin.
This paper describes a method of inflecting and linearizing a lemmatized dependency tree by: (1) determining a regular expression and substitution to describe each productive wordform rule; (2) learning the dependency distance tolerance for each head-dependent pair, resulting in an edge-weighted directed acyclic graph (DAG); and (3) topologically sorting the DAG into a surface realization based on edge weight. The method’s output for 11 languages across 18 treebanks is competitive with the other submissions to the Second Multilingual Surface Realization Shared Task (SR ‘19).
The Surface Realization Shared Task involves mapping Universal Dependency graphs to raw text, i.e. restoring word order and inflection from a graph of typed, directed dependencies between lemmas. Interpreted Regular Tree Grammars (IRTGs) encode the correspondence between generations in multiple algebras, and have previously been used for semantic parsing from raw text. Our system induces an IRTG for simultaneously building pairs of surface forms and UD graphs in the SRST training data, then prunes this grammar for each UD graph in the test data for efficient parsing and generation of the surface ordering of lemmas. For the inflection step we use a standard sequence-to-sequence model with a biLSTM encoder and an LSTM decoder with attention. Both components of our system are available on GitHub under an MIT license.
We first describe a surface realizer forUniversal Dependencies (UD) structures. The system uses a symbolic approach to transform the dependency tree into a tree of constituents that is transformed into an English sentence by an existing realizer. This approach was then adapted for the two shared tasks of SR’19. The system is quite fast and showed competitive results for English sentences using automatic and manual evaluation measures.
We introduce the IMS contribution to the Surface Realization Shared Task 2019. Our submission achieves the state-of-the-art performance without using any external resources. The system takes a pipeline approach consisting of five steps: linearization, completion, inflection, contraction, and detokenization. We compare the performance of our linearization algorithm with two external baselines and report results for each step in the pipeline. Furthermore, we perform detailed error analysis revealing correlation between word order freedom and difficulty of the linearization task.
This study describes the approach developed by the Tilburg University team to the shallow track of the Multilingual Surface Realization Shared Task 2019 (SR’19) (Mille et al., 2019). Based on Ferreira et al. (2017) and on our 2018 submission Ferreira et al. (2018), the approach generates texts by first preprocessing an input dependency tree into an ordered linearized string, which is then realized using a rule-based and a statistical machine translation (SMT) model. This year our submission is able to realize texts in the 11 languages proposed for the task, different from our last year submission, which covered only 6 Indo-European languages. The model is publicly available.
This paper presents the model we developed for the shallow track of the 2019 NLG Surface Realization Shared Task. The model reconstructs sentences whose word order and word inflections were removed. We divided the problem into two sub-problems: reordering and inflecting. For the purpose of reordering, we used a pointer network integrated with a transformer model as its encoder-decoder modules. In order to generate the inflected forms of tokens, a Feed Forward Neural Network was employed.
We describe our exploratory system for the shallow surface realization task, which combines morphological inflection using character sequence-to-sequence models with a baseline linearizer that implements a tree-to-tree model using sequence-to-sequence models on serialized trees. Results for morphological inflection were competitive across languages. Due to time constraints, we could only submit complete results (including linearization) for English. Preliminary linearization results were decent, with a small benefit from reranking to prefer valid output trees, but inadequate control over the words in the output led to poor quality on longer sentences.
The Multilingual Surface Realization Shared Task 2019 focuses on generating sentences from lemmatized sets of universal dependency parses with rich features. This paper describes the results of our participation in the deep track. The core innovation in our approach is to use a graph convolutional network to encode the dependency trees given as input. Upon adding morphological features, our system achieves the third rank without using data augmentation techniques or additional components (such as a re-ranker).
We describe the system presented at the SR’19 shared task by the DipInfoUnito team. Our approach is based on supervised machine learning. In particular, we divide the SR task into two independent subtasks, namely word order prediction and morphology inflection prediction. Two neural networks with different architectures run on the same input structure, each producing a partial output which is recombined in the final step in order to produce the predicted surface form. This work is a direct successor of the architecture presented at SR’19.
This paper presents the LORIA / Lorraine University submission at the Multilingual Surface Realisation shared task 2019 for the shallow track. We outline our approach and evaluate it on 11 languages covered by the shared task. We provide a separate evaluation of each component of our pipeline, concluding on some difficulties and suggesting directions for future work.
This paper presents an exploratory study that aims to evaluate the usefulness of back-translation in Natural Language Generation (NLG) from semantic representations for non-English languages. Specifically, Abstract Meaning Representation and Brazilian Portuguese (BP) are chosen as semantic representation and language, respectively. Two methods (focused on Statistical and Neural Machine Translation) are evaluated on two datasets (one automatically generated and another one human-generated) to compare the performance in a real context. Also, several cuts according to quality measures are performed to evaluate the importance (or not) of the data quality in NLG. Results show that there are still many improvements to be made but this is a promising approach.
Neural Module Networks, originally proposed for the task of visual question answering, are a class of neural network architectures that involve human-specified neural modules, each designed for a specific form of reasoning. In current formulations of such networks only the parameters of the neural modules and/or the order of their execution is learned. In this work, we further expand this approach and also learn the underlying internal structure of modules in terms of the ordering and combination of simple and elementary arithmetic operators. We utilize a minimum amount of prior knowledge from the human-specified neural modules in the form of different input types and arithmetic operators used in these modules. Our results show that one is indeed able to simultaneously learn both internal module structure and module sequencing without extra supervisory signals for module execution sequencing. With this approach, we report performance comparable to models using hand-designed modules. In addition, we do a analysis of sensitivity of the learned modules w.r.t. the arithmetic operations and infer the analytical expressions of the learned modules.
In this paper, we propose a new approach to learn multimodal multilingual embeddings for matching images and their relevant captions in two languages. We combine two existing objective functions to make images and captions close in a joint embedding space while adapting the alignment of word embeddings between existing languages in our model. We show that our approach enables better generalization, achieving state-of-the-art performance in text-to-image and image-to-text retrieval task, and caption-caption similarity task. Two multimodal multilingual datasets are used for evaluation: Multi30k with German and English captions and Microsoft-COCO with English and Japanese captions.
In this paper, we experiment with a recently proposed visual reasoning task dealing with quantities – modeling the multimodal, contextually-dependent meaning of size adjectives (‘big’, ‘small’) – and explore the impact of varying the training data on the learning behavior of a state-of-art system. In previous work, models have been shown to fail in generalizing to unseen adjective-noun combinations. Here, we investigate whether, and to what extent, seeing some of these cases during training helps a model understand the rule subtending the task, i.e., that being big implies being not small, and vice versa. We show that relatively few examples are enough to understand this relationship, and that developing a specific, mutually exclusive representation of size adjectives is beneficial to the task.
Chinese characters are unique in its logographic nature, which inherently encodes world knowledge through thousands of years evolution. This paper proposes an embedding approach, namely eigencharacter (EC) space, which helps NLP application easily access the knowledge encoded in Chinese orthography. These EC representations are automatically extracted, encode both structural and radical information, and easily integrate with other computational models. We built EC representations of 5,000 Chinese characters, investigated orthography knowledge encoded in ECs, and demonstrated how these ECs identified visually similar characters with both structural and radical information.
Scene graphs represent semantic information in images, which can help image captioning system to produce more descriptive outputs versus using only the image as context. Recent captioning approaches rely on ad-hoc approaches to obtain graphs for images. However, those graphs introduce noise and it is unclear the effect of parser errors on captioning accuracy. In this work, we investigate to what extent scene graphs can help image captioning. Our results show that a state-of-the-art scene graph parser can boost performance almost as much as the ground truth graphs, showing that the bottleneck currently resides more on the captioning models than on the performance of the scene graph parser.
It is assumed that multimodal machine translation systems are better than text-only systems at translating phrases that have a direct correspondence in the image. This assumption has been challenged in experiments demonstrating that state-of-the-art multimodal systems perform equally well in the presence of randomly selected images, but, more recently, it has been shown that masking entities from the source language sentence during training can help to overcome this problem. In this paper, we conduct experiments with both visual and textual adversaries in order to understand the role of incorrect textual inputs to such systems. Our results show that when the source language sentence contains mistakes, multimodal translation systems do not leverage the additional visual signal to produce the correct translation. We also find that the degradation of translation performance caused by textual adversaries is significantly higher than by visual adversaries.
Previous research into agent communication has shown that a pre-trained guide can speed up the learning process of an imitation learning agent. The guide achieves this by providing the agent with discrete messages in an emerged language about how to solve the task. We extend this one-directional communication by a one-bit communication channel from the learner back to the guide: It is able to ask the guide for help, and we limit the guidance by penalizing the learner for these requests. During training, the agent learns to control this gate based on its current observation. We find that the amount of requested guidance decreases over time and guidance is requested in situations of high uncertainty. We investigate the agent’s performance in cases of open and closed gates and discuss potential motives for the observed gating behavior.
Readers’ eye movements used as part of the training signal have been shown to improve performance in a wide range of Natural Language Processing (NLP) tasks. Previous work uses gaze data either at the type level or at the token level and mostly from a single eye-tracking corpus. In this paper, we analyze type vs token-level integration options with eye tracking data from two corpora to inform two syntactic sequence labeling problems: binary phrase chunking and part-of-speech tagging. We show that using globally-aggregated measures that capture the central tendency or variability of gaze data is more beneficial than proposed local views which retain individual participant information. While gaze data is informative for supervised POS tagging, which complements previous findings on unsupervised POS induction, almost no improvement is obtained for binary phrase chunking, except for a single specific setup. Hence, caution is warranted when using gaze data as signal for NLP, as no single view is robust over tasks, modeling choice and gaze corpus.
How can we teach artificial agents to use human language flexibly to solve problems in real-world environments? We have an example of this in nature: human babies eventually learn to use human language to solve problems, and they are taught with an adult human-in-the-loop. Unfortunately, current machine learning methods (e.g. from deep reinforcement learning) are too data inefficient to learn language in this way. An outstanding goal is finding an algorithm with a suitable ‘language learning prior’ that allows it to learn human language, while minimizing the number of on-policy human interactions. In this paper, we propose to learn such a prior in simulation using an approach we call, Learning to Learn to Communicate (L2C). Specifically, in L2C we train a meta-learning agent in simulation to interact with populations of pre-trained agents, each with their own distinct communication protocol. Once the meta-learning agent is able to quickly adapt to each population of agents, it can be deployed in new populations, including populations speaking human language. Our key insight is that such populations can be obtained via self-play, after pre-training agents with imitation learning on a small amount of off-policy human language data. We call this latter technique Seeded Self-Play (S2P). Our preliminary experiments show that agents trained with L2C and S2P need fewer on-policy samples to learn a compositional language in a Lewis signaling game.
We present the results of the second Fact Extraction and VERification (FEVER2.0) Shared Task. The task challenged participants to both build systems to verify factoid claims using evidence retrieved from Wikipedia and to generate adversarial attacks against other participant’s systems. The shared task had three phases: building, breaking and fixing. There were 8 systems in the builder’s round, three of which were new qualifying submissions for this shared task, and 5 adversaries generated instances designed to induce classification errors and one builder submitted a fixed system which had higher FEVER score and resilience than their first submission. All but one newly submitted systems attained FEVER scores higher than the best performing system from the first shared task and under adversarial evaluation, all systems exhibited losses in FEVER score. There was a great variety in adversarial attack types as well as the techniques used to generate the attacks, In this paper, we present the results of the shared task and a summary of the systems, highlighting commonalities and innovations among participating systems.
The goal of our paper is to compare psycholinguistic text features with fact checking approaches to distinguish lies from true statements. We examine both methods using data from a large ongoing study on deception and deception detection covering a mixture of factual and opinionated topics that polarize public opinion. We conclude that fact checking approaches based on Wikipedia are too limited for this task, as only a few percent of sentences from our study has enough evidence to become supported or refuted. Psycholinguistic features turn out to outperform both fact checking and human baselines, but the accuracy is not high. Overall, it appears that deception detection applicable to less-than-obvious topics is a difficult task and a problem to be solved.
We present a multi-task learning model that leverages large amount of textual information from existing datasets to improve stance prediction. In particular, we utilize multiple NLP tasks under both unsupervised and supervised settings for the target stance prediction task. Our model obtains state-of-the-art performance on a public benchmark dataset, Fake News Challenge, outperforming current approaches by a wide margin.
We present our Generative Enhanced Model (GEM) that we used to create samples awarded the first prize on the FEVER 2.0 Breakers Task. GEM is the extended language model developed upon GPT-2 architecture. The addition of novel target vocabulary input to the already existing context input enabled controlled text generation. The training procedure resulted in creating a model that inherited the knowledge of pretrained GPT-2, and therefore was ready to generate natural-like English sentences in the task domain with some additional control. As a result, GEM generated malicious claims that mixed facts from various articles, so it became difficult to classify their truthfulness.
In this paper, we propose a new approach to learn multimodal multilingual embeddings for matching images and their relevant captions in two languages. We combine two existing objective functions to make images and captions close in a joint embedding space while adapting the alignment of word embeddings between existing languages in our model. We show that our approach enables better generalization, achieving state-of-the-art performance in text-to-image and image-to-text retrieval task, and caption-caption similarity task. Two multimodal multilingual datasets are used for evaluation: Multi30k with German and English captions and Microsoft-COCO with English and Japanese captions.
The recent demonstration of the power of huge language models such as GPT-2 to memorise the answers to factoid questions raises questions about the extent to which knowledge is being embedded directly within these large models. This short paper describes an architecture through which much smaller models can also answer such questions - by making use of ‘raw’ external knowledge. The contribution of this work is that the methods presented here rely on unsupervised learning techniques, complementing the unsupervised training of the Language Model. The goal of this line of research is to be able to add knowledge explicitly, without extensive training.
We present a scalable, open-source platform that “distills” a potentially large text collection into a knowledge graph. Our platform takes documents stored in Apache Solr and scales out the Stanford CoreNLP toolkit via Apache Spark integration to extract mentions and relations that are then ingested into the Neo4j graph database. The raw knowledge graph is then enriched with facts extracted from an external knowledge graph. The complete product can be manipulated by various applications using Neo4j’s native Cypher query language: We present a subgraph-matching approach to align extracted relations with external facts and show that fact verification, locating textual support for asserted facts, detecting inconsistent and missing facts, and extracting distantly-supervised training data can all be performed within the same framework.
Many previous studies on relation extrac-tion have been focused on finding only one relation between two entities in a single sentence. However, we can easily find the fact that multiple entities exist in a single sentence and the entities form multiple relations. To resolve this prob-lem, we propose a relation extraction model based on a dual pointer network with a multi-head attention mechanism. The proposed model finds n-to-1 subject-object relations by using a forward de-coder called an object decoder. Then, it finds 1-to-n subject-object relations by using a backward decoder called a sub-ject decoder. In the experiments with the ACE-05 dataset and the NYT dataset, the proposed model achieved the state-of-the-art performances (F1-score of 80.5% in the ACE-05 dataset, F1-score of 78.3% in the NYT dataset)
Recent Deep Learning (DL) models have succeeded in achieving human-level accuracy on various natural language tasks such as question-answering, natural language inference (NLI), and textual entailment. These tasks not only require the contextual knowledge but also the reasoning abilities to be solved efficiently. In this paper, we propose an unsupervised question-answering based approach for a similar task, fact-checking. We transform the FEVER dataset into a Cloze-task by masking named entities provided in the claims. To predict the answer token, we utilize pre-trained Bidirectional Encoder Representations from Transformers (BERT). The classifier computes label based on the correctly answered questions and a threshold. Currently, the classifier is able to classify the claims as “SUPPORTS” and “MANUAL_REVIEW”. This approach achieves a label accuracy of 80.2% on the development set and 80.25% on the test set of the transformed dataset.
Recognizing the implicit link between a claim and a piece of evidence (i.e. warrant) is the key to improving the performance of evidence detection. In this work, we explore the effectiveness of automatically extracted warrants for evidence detection. Given a claim and candidate evidence, our proposed method extracts multiple warrants via similarity search from an existing, structured corpus of arguments. We then attentively aggregate the extracted warrants, considering the consistency between the given argument and the acquired warrants. Although a qualitative analysis on the warrants shows that the extraction method needs to be improved, our results indicate that our method can still improve the performance of evidence detection.
One of the important tasks in opinion mining is to extract aspects of the opinion target. Aspects are features or characteristics of the opinion target that are being reviewed, which can be categorised into explicit and implicit aspects. Extracting aspects from opinions is essential in order to ensure accurate information about certain attributes of an opinion target is retrieved. For instance, a professional camera receives a positive feedback in terms of its functionalities in a review, but its overly high price receives negative feedback. Most of the existing solutions focus on explicit aspects. However, sentences in reviews normally do not state the aspects explicitly. In this research, two hybrid models are proposed to identify and extract both explicit and implicit aspects, namely TDM-DC and TDM-TED. The proposed models combine topic modelling and dictionary-based approach. The models are unsupervised as they do not require any labelled dataset. The experimental results show that TDM-DC achieves F1-measure of 58.70%, where it outperforms both the baseline topic model and dictionary-based approach. In comparison to other existing unsupervised techniques, the proposed models are able to achieve higher F1-measure by approximately 3%. Although the supervised techniques perform slightly better, the proposed models are domain-independent, and hence more versatile.
Triggered by Internet development, a large amount of information is published in online sources. However, it is a well-known fact that publications are inundated with inaccurate data. That is why fact-checking has become a significant topic in the last 5 years. It is widely accepted that factual data verification is a challenge even for the experts. This paper presents a domain-independent fact checking system. It can solve the fact verification problem entirely or at the individual stages. The proposed model combines various advanced methods of text data analysis, such as BERT and Infersent. The theoretical and empirical study of the system features is carried out. Based on FEVER and Fact Checking Challenge test-collections, experimental results demonstrate that our model can achieve the score on a par with state-of-the-art models designed by the specificity of particular datasets.
Finding evidence is of vital importance in research as well as fact checking and an evidence detection method would be useful in speeding up this process. However, when addressing a new topic there is no training data and there are two approaches to get started. One could use large amounts of out-of-domain data to train a state-of-the-art method, or to use the small data that a person creates while working on the topic. In this paper, we address this problem in two steps. First, by simulating users who read source documents and label sentences they can use as evidence, thereby creating small amounts of training data for an interactively trained evidence detection model; and second, by comparing such an interactively trained model against a pre-trained model that has been trained on large out-of-domain data. We found that an interactively trained model not only often out-performs a state-of-the-art model but also requires significantly lower amounts of computational resources. Therefore, especially when computational resources are scarce, e.g. no GPU available, training a smaller model on the fly is preferable to training a well generalising but resource hungry out-of-domain model.
Defined as the intentional or unintentionalspread of false information (K et al., 2019)through context and/or content manipulation,fake news has become one of the most seriousproblems associated with online information(Waldrop, 2017). Consequently, it comes asno surprise that Fake News Detection hasbecome one of the major foci of variousfields of machine learning and while machinelearning models have allowed individualsand companies to automate decision-basedprocesses that were once thought to be onlydoable by humans, it is no secret that thereal-life applications of such models are notviable without the existence of an adequatetraining dataset. In this paper we describethe Veritas Annotator, a web application formanually identifying the origin of a rumour.These rumours, often referred as claims,were previously checked for validity byFact-Checking Agencies.
We describe our submission for the Breaker phase of the second Fact Extraction and VERification (FEVER) Shared Task. Our adversarial data can be explained by two perspectives. First, we aimed at testing model’s ability to retrieve evidence, when appropriate query terms could not be easily generated from the claim. Second, we test model’s ability to precisely understand the implications of the texts, which we expect to be rare in FEVER 1.0 dataset. Overall, we suggested six types of adversarial attacks. The evaluation on the submitted systems showed that the systems were only able get both the evidence and label correct in 20% of the data. We also demonstrate our adversarial run analysis in the data development process.
This paper contains our system description for the second Fact Extraction and VERification (FEVER) challenge. We propose a two-staged sentence selection strategy to account for examples in the dataset where evidence is not only conditioned on the claim, but also on previously retrieved evidence. We use a publicly available document retrieval module and have fine-tuned BERT checkpoints for sentence se- lection and as the entailment classifier. We report a FEVER score of 68.46% on the blind testset.
Fever Shared 2.0 Task is a challenge meant for developing automated fact checking systems. Our approach for the Fever 2.0 is based on a previous proposal developed by Team Athene UKP TU Darmstadt. Our proposal modifies the sentence retrieval phase, using statement extraction and representation in the form of triplets (subject, object, action). Triplets are extracted from the claim and compare to triplets extracted from Wikipedia articles using semantic similarity. Our results are satisfactory but there is room for improvement.
Much work in contemporary computational semantics follows the distributional hypothesis (DH), which is understood as an approach to semantics according to which the meaning of a word is a function of its distribution over contexts which is represented as vectors (word embeddings) within a multi-dimensional semantic space. In practice, use is identified with occurrence in text corpora, though there are some efforts to use corpora containing multi-modal information. In this paper we argue that the distributional hypothesis is intrinsically misguided as a self-supporting basis for semantics, as Firth was entirely aware. We mention philosophical arguments concerning the lack of normativity within DH data. Furthermore, we point out the shortcomings of DH as a model of learning, by discussing a variety of linguistic classes that cannot be learnt on a distributional basis, including indexicals, proper names, and wh-phrases. Instead of pursuing DH, we sketch an account of the problematic learning cases by integrating a rich, Firthian notion of dialogue context with interactive learning in signalling games backed by in probabilistic Type Theory with Records. We conclude that the success of the DH in computational semantics rests on a post hoc effect: DS presupposes a referential semantics on the basis of which utterances can be produced, comprehended and analysed in the first place.
Generating “commonsense’’ knowledge for intelligent understanding and reasoning is a difficult, long-standing problem, whose scale challenges the capacity of any approach driven primarily by human input. Furthermore, approaches based on mining statistically repetitive patterns fail to produce the rich representations humans acquire, and fall far short of human efficiency in inducing knowledge from text. The idea of our approach to this problem is to provide a learning system with a “head start” consisting of a semantic parser, some basic ontological knowledge, and most importantly, a small set of very general schemas about the kinds of patterns of events (often purposive, causal, or socially conventional) that even a one- or two-year-old could reasonably be presumed to possess. We match these initial schemas to simple children’s stories, obtaining concrete instances, and combining and abstracting these into new candidate schemas. Both the initial and generated schemas are specified using a rich, expressive logical form. While modern approaches to schema reasoning often only use slot-and-filler structures, this logical form allows us to specify complex relations and constraints over the slots. Though formal, the representations are language-like, and as such readily relatable to NL text. The agents, objects, and other roles in the schemas are represented by typed variables, and the event variables can be related through partial temporal ordering and causal relations. To match natural language stories with existing schemas, we first parse the stories into an underspecified variant of the logical form used by the schemas, which is suitable for most concrete stories. We include a walkthrough of matching a children’s story to these schemas and generating inferences from these matches.
Dependent Type Semantics (DTS; Bekki and Mineshima, 2017) is a proof-theoretic compositional dynamic semantics based on Dependent Type Theory. The semantic representations for declarative sentences in DTS are types, based on the propositions-as-types paradigm. While type-theoretic semantics for natural language based on dependent type theory has been developed by many authors, how to assign semantic representations to interrogative sentences has been a non-trivial problem. In this study, we show how to provide the semantics of interrogative sentences in DTS. The basic idea is to assign the same type to both declarative sentences and interrogative sentences, partly building on the recent proposal in Inquisitive Semantics. We use Combinatory Categorial Grammar (CCG) as a syntactic component of DTS and implement our compositional semantics for interrogative sentences using ccg2lambda, a semantic parsing platform based on CCG. Based on the idea that the relationship between questions and answers can be formulated as the task of Recognizing Textual Entailment (RTE), we implement our inference system using proof assistant Coq and show that our system can deal with a wide range of question-answer relationships discussed in the formal semantics literature, including those with polar questions, alternative questions, and wh-questions.
We outline a hyperintensional situation semantics in which hyperintensionality is modelled as a ‘side effect’, as this term has been understood in natural language semantics and in functional programming. We use monads from category theory in order to ‘upgrade’ an ordinary intensional semantics to a possible hyperintensional counterpart.
Claims database and electronic health records database do not usually capture kinship or family relationship information, which is imperative for genetic research. We identify online obituaries as a new data source and propose a special named entity recognition and relation extraction solution to extract names and kinships from online obituaries. Built on 1,809 annotated obituaries and a novel tagging scheme, our joint neural model achieved macro-averaged precision, recall and F measure of 72.69%, 78.54% and 74.93%, and micro-averaged precision, recall and F measure of 95.74%, 98.25% and 96.98% using 57 kinships with 10 or more examples in a 10-fold cross-validation experiment. The model performance improved dramatically when trained with 34 kinships with 50 or more examples. Leveraging additional information such as age, death date, birth date and residence mentioned by obituaries, we foresee a promising future of supplementing EHR databases with comprehensive and accurate kinship information for genetic research.
In the medical domain, user-generated social media text is increasingly used as a valuable complementary knowledge source to scientific medical literature. The extraction of this knowledge is complicated by colloquial language use and misspellings. Yet, lexical normalization of such data has not been addressed properly. This paper presents an unsupervised, data-driven spelling correction module for medical social media. Our method outperforms state-of-the-art spelling correction and can detect mistakes with an F0.5 of 0.888. Additionally, we present a novel corpus for spelling mistake detection and correction on a medical patient forum.
The number of users of social media continues to grow, with nearly half of adults worldwide and two-thirds of all American adults using social networking. Advances in automated data processing, machine learning and NLP present the possibility of utilizing this massive data source for biomedical and public health applications, if researchers address the methodological challenges unique to this media. We present the Social Media Mining for Health Shared Tasks collocated with the ACL at Florence in 2019, which address these challenges for health monitoring and surveillance, utilizing state of the art techniques for processing noisy, real-world, and substantially creative language expressions from social media users. For the fourth execution of this challenge, we proposed four different tasks. Task 1 asked participants to distinguish tweets reporting an adverse drug reaction (ADR) from those that do not. Task 2, a follow-up to Task 1, asked participants to identify the span of text in tweets reporting ADRs. Task 3 is an end-to-end task where the goal was to first detect tweets mentioning an ADR and then map the extracted colloquial mentions of ADRs in the tweets to their corresponding standard concept IDs in the MedDRA vocabulary. Finally, Task 4 asked participants to classify whether a tweet contains a personal mention of one’s health, a more general discussion of the health issue, or is an unrelated mention. A total of 34 teams from around the world registered and 19 teams from 12 countries submitted a system run. We summarize here the corpora for this challenge which are freely available at https://competitions.codalab.org/competitions/22521, and present an overview of the methods and the results of the competing systems.
The medical concept normalisation task aims to map textual descriptions to standard terminologies such as SNOMED-CT or MedDRA. Existing publicly available datasets annotated using different terminologies cannot be simply merged and utilised, and therefore become less valuable when developing machine learning-based concept normalisation systems. To address that, we designed a data harmonisation pipeline and engineered a corpus of 27,979 textual descriptions simultaneously mapped to both MedDRA and SNOMED-CT, sourced from five publicly available datasets across biomedical and social media domains. The pipeline can be used in the future to integrate new datasets into the corpus and also could be applied in relevant data curation tasks. We also described a method to merge different terminologies into a single concept graph preserving their relations and demonstrated that representation learning approach based on random walks on a graph can efficiently encode both hierarchical and equivalent relations and capture semantic similarities not only between concepts inside a given terminology but also between concepts from different terminologies. We believe that making a corpus and embeddings for cross-terminology medical concept normalisation available to the research community would contribute to a better understanding of the task.
Depression and anxiety are the two most prevalent mental health disorders worldwide, impacting the lives of millions of people each year. In this work, we develop and evaluate a multilabel, multidimensional deep neural network designed to predict PHQ-4 scores based on individuals written text. Our system outperforms random baseline metrics and provides a novel approach to how we can predict psychometric scores from written text. Additionally, we explore how this architecture can be applied to analyse social media data.
This is the system description of the Harbin Institute of Technology Shenzhen (HITSZ) team for the first and second subtasks of the fourth Social Media Mining for Health Applications (SMM4H) shared task in 2019. The two subtasks are automatic classification and extraction of adverse effect mentions in tweets. The systems for the two subtasks are based on bidirectional encoder representations from transformers (BERT), and achieves promising results. Among the systems we developed for subtask1, the best F1-score was 0.6457, for subtask2, the best relaxed F1-score and the best strict F1-score were 0.614 and 0.407 respectively. Our system ranks first among all systems on subtask1.
This paper describes a system developed for the Social Media Mining for Health (SMM4H) 2019 shared tasks. Specifically, we participated in three tasks. The goals of the first two tasks are to classify whether a tweet contains mentions of adverse drug reactions (ADR) and extract these mentions, respectively. The objective of the third task is to build an end-to-end solution: first, detect ADR mentions and then map these entities to concepts in a controlled vocabulary. We investigate the use of a language representation model BERT trained to obtain semantic representations of social media texts. Our experiments on a dataset of user reviews showed that BERT is superior to state-of-the-art models based on recurrent neural networks. The BERT-based system for Task 1 obtained an F1 of 57.38%, with improvements up to +7.19% F1 over a score averaged across all 43 submissions. The ensemble of neural networks with a voting scheme for named entity recognition ranked first among 9 teams at the SMM4H 2019 Task 2 and obtained a relaxed F1 of 65.8%. The end-to-end model based on BERT for ADR normalization ranked first at the SMM4H 2019 Task 3 and obtained a relaxed F1 of 43.2%.
We describe our submissions to the 4th edition of the Social Media Mining for Health Applications (SMM4H) shared task. Our team (UZH) participated in two sub-tasks: Automatic classifications of adverse effects mentions in tweets (Task 1) and Generalizable identification of personal health experience mentions (Task 4). For our submissions, we exploited ensembles based on a pre-trained language representation with a neural transformer architecture (BERT) (Tasks 1 and 4) and a CNN-BiLSTM(-CRF) network within a multi-task learning scenario (Task 1). These systems are placed on top of a carefully crafted pipeline of domain-specific preprocessing steps.
Identifying mentions of medical concepts in social media is challenging because of high variability in free text. In this paper, we propose a novel neural network architecture, the Collocated LSTM with Attentive Pooling and Aggregated representation (CLAPA), that integrates a bidirectional LSTM model with attention and pooling strategy and utilizes the collocation information from training data to improve the representation of medical concepts. The collocation and aggregation layers improve the model performance on the task of identifying mentions of adverse drug events (ADE) in tweets. Using the dataset made available as part of the workshop shared task, we show that careful selection of neighborhood contexts can help uncover useful local information and improve the overall medical concept representation.
We study how language on social media is linked to mortal diseases such as atherosclerotic heart disease (AHD), diabetes and various types of cancer. Our proposed model leverages state-of-the-art sentence embeddings, followed by a regression model and clustering, without the need of additional labelled data. It allows to predict community-level medical outcomes from language, and thereby potentially translate these to the individual level. The method is applicable to a wide range of target variables and allows us to discover known and potentially novel correlations of medical outcomes with life-style aspects and other socioeconomic risk factors.
The increase in the prevalence of mental health problems has coincided with a growing popularity of health related social networking sites. Regardless of their therapeutic potential, on-line support groups (OSGs) can also have negative effects on patients. In this work we propose a novel methodology to automatically verify the presence of therapeutic factors in social networking websites by using Natural Language Processing (NLP) techniques. The methodology is evaluated on on-line asynchronous multi-party conversations collected from an OSG and Twitter. The results of the analysis indicate that therapeutic factors occur more frequently in OSG conversations than in Twitter conversations. Moreover, the analysis of OSG conversations reveals that the users of that platform are supportive, and interactions are likely to lead to the improvement of their emotional state. We believe that our method provides a stepping stone towards automatic analysis of emotional states of users of online platforms. Possible applications of the method include provision of guidelines that highlight potential implications of using such platforms on users’ mental health, and/or support in the analysis of their impact on specific individuals.
Transfer learning is promising for many NLP applications, especially in tasks with limited labeled data. This paper describes the methods developed by team TMRLeiden for the 2019 Social Media Mining for Health Applications (SMM4H) Shared Task. Our methods use state-of-the-art transfer learning methods to classify, extract and normalise adverse drug effects (ADRs) and to classify personal health mentions from health-related tweets. The code and fine-tuned models are publicly available.
This paper describes a system for automatically classifying adverse effects mentions in tweets developed for the task 1 at Social Media Mining for Health Applications (SMM4H) Shared Task 2019. We have developed a system based on LSTM neural networks inspired by the excellent results obtained by deep learning classifiers in the last edition of this task. The network is trained along with Twitter GloVe pre-trained word embeddings.
This paper describes our system for the first and second shared tasks of the fourth Social Media Mining for Health Applications (SMM4H) workshop. We enhance tweet representation with a language model and distinguish the importance of different words with Multi-Head Self-Attention. In addition, transfer learning is exploited to make up for the data shortage. Our system achieved competitive results on both tasks with an F1-score of 0.5718 for task 1 and 0.653 (overlap) / 0.357 (strict) for task 2.
Social Media Mining for Health Applications (SMM4H) Adverse Effect Mentions Shared Task challenges participants to accurately identify spans of text within a tweet that correspond to Adverse Effects (AEs) resulting from medication usage (Weissenbacher et al., 2019). This task features a training data set of 2,367 tweets, in addition to a 1,000 tweet evaluation data set. The solution presented here features a bidirectional Long Short-term Memory Network (bi-LSTM) for the generation of character-level embeddings. It uses a second bi-LSTM trained on both character and token level embeddings to feed a Conditional Random Field (CRF) which provides the final classification. This paper further discusses the deep learning algorithms used in our solution.
Over time the use of social networks is becoming very popular platforms for sharing health related information. Social Media Mining for Health Applications (SMM4H) provides tasks such as those described in this document to help manage information in the health domain. This document shows the first participation of the SINAI group. We study approaches based on machine learning and deep learning to extract adverse drug reaction mentions from Twitter. The results obtained in the tasks are encouraging, we are close to the average of all participants and even above in some cases.
We participated in Task 1 of the Social Media Mining for Health Applications (SMM4H) 2019 Shared Tasks on detecting mentions of adverse drug events (ADEs) in tweets. Our approach relied on a text processing pipeline for tweets, and training traditional machine learning and deep learning models. Our submitted runs performed above average for the task.
This paper describes the system developed by team ASU-NLP for the Social Media Mining for Health Applications(SMM4H) shared task 4. We extract feature embeddings from the BioBERT (Lee et al., 2019) model which has been fine-tuned on the training dataset and use that as inputs to a dense fully connected neural network. We achieve above average scores among the participant systems with the overall F1-score, accuracy, precision, recall as 0.8036, 0.8456, 0.9783, 0.6818 respectively.
This paper describes the system that team MYTOMORROWS-TU DELFT developed for the 2019 Social Media Mining for Health Applications (SMM4H) Shared Task 3, for the end-to-end normalization of ADR tweet mentions to their corresponding MEDDRA codes. For the first two steps, we reuse a state-of-the art approach, focusing our contribution on the final entity-linking step. For that we propose a simple Few-Shot learning approach, based on pre-trained word embeddings and data from the UMLS, combined with the provided training data. Our system (relaxed F1: 0.337-0.345) outperforms the average (relaxed F1 0.2972) of the participants in this task, demonstrating the potential feasibility of few-shot learning in the context of medical text normalization.
In this study, we describe our methods to automatically classify Twitter posts conveying events of adverse drug reaction (ADR). Based on our previous experience in tackling the ADR classification task, we empirically applied the vote-based under-sampling ensemble approach along with linear support vector machine (SVM) to develop our classifiers as part of our participation in ACL 2019 Social Media Mining for Health Applications (SMM4H) shared task 1. The best-performed model on the test sets were trained on a merged corpus consisting of the datasets released by SMM4H 2017 and 2019. By using VUE, the corpus was randomly under-sampled with 2:1 ratio between the negative and positive classes to create an ensemble using the linear kernel trained with features including bag-of-word, domain knowledge, negation and word embedding. The best performing model achieved an F-measure of 0.551 which is about 5% higher than the average F-scores of 16 teams.
This paper describes the models used by our team in SMM4H 2019 shared task. We submitted results for subtasks 1 and 2. For task 1 which aims to detect tweets with Adverse Drug Reaction (ADR) mentions we used ELMo embeddings which is a deep contextualized word representation able to capture both syntactic and semantic characteristics. For task 2, which focuses on extraction of ADR mentions, first the same architecture as task 1 was used to identify whether or not a tweet contains ADR. Then, for tweets positively classified as mentioning ADR, the relevant text span was identified by similarity matching with 3 different lexicon sets.
CLaC labs participated in Task 1 and 4 of SMM4H 2019. We pursed two main objectives in our submission. First we tried to use some textual features in a deep net framework, and second, the potential use of more than one word embedding was tested. The results seem positively affected by the proposed architectures.
In this paper, we present our approach and the system description for the Social Media Mining for Health Applications (SMM4H) Shared Task 1,2 and 4 (2019). Our main contribution is to show the effectiveness of Transfer Learning approaches like BERT and ULMFiT, and how they generalize for the classification tasks like identification of adverse drug reaction mentions and reporting of personal health problems in tweets. We show the use of stacked embeddings combined with BLSTM+CRF tagger for identifying spans mentioning adverse drug reactions in tweets. We also show that these approaches perform well even with imbalanced dataset in comparison to undersampling and oversampling.
This paper details our approach to the task of detecting reportage of adverse drug reaction in tweets as part of the 2019 social media mining for healthcare applications shared task. We employed a combination of three types of word representations as input to a LSTM model. With this approach, we achieved an F1 score of 0.5209.
Analyzing social media posts can offer insights into a wide range of topics that are commonly discussed online, providing valuable information for studying various health-related phenomena reported online. The outcome of this work can offer insights into pharmacovigilance research to monitor the adverse effects of medications. This research specifically looks into mentions of adverse drug reactions (ADRs) in Twitter data through the Social Media Mining for Health Applications (SMM4H) Shared Task 2019. Adverse drug reactions are undesired harmful effects which can arise from medication or other methods of treatment. The goal of this research is to build accurate models using natural language processing techniques to detect reports of adverse drug reactions in Twitter data and extract these words or phrases.
This work attempts to explain the types of computation that neural networks can perform by relating them to automata. We first define what it means for a real-time network with bounded precision to accept a language. A measure of network memory follows from this definition. We then characterize the classes of languages acceptable by various recurrent networks, attention, and convolutional networks. We find that LSTMs function like counter machines and relate convolutional networks to the subregular hierarchy. Overall, this work attempts to increase our understanding and ability to interpret neural networks through the lens of theory. These theoretical insights help explain neural computation, as well as the relationship between neural networks and natural language grammar.
While sequence-to-sequence (seq2seq) models achieve state-of-the-art performance in many natural language processing tasks, they can be too slow for real-time applications. One performance bottleneck is predicting the most likely next token over a large vocabulary; methods to circumvent this bottleneck are a current research topic. We focus specifically on using seq2seq models for semantic parsing, where we observe that grammars often exist which specify valid formal representations of utterance semantics. By developing a generic approach for restricting the predictions of a seq2seq model to grammatically permissible continuations, we arrive at a widely applicable technique for speeding up semantic parsing. The technique leads to a 74% speed-up on an in-house dataset with a large vocabulary, compared to the same neural model without grammatical restrictions
We analyse Recurrent Neural Networks (RNNs) to understand the significance of multiple LSTM layers. We argue that the Weighted Finite-state Automata (WFA) trained using a spectral learning algorithm are helpful to analyse RNNs. Our results suggest that multiple LSTM layers in RNNs help learning distributed hidden states, but have a smaller impact on the ability to learn long-term dependencies. The analysis is based on the empirical results, however relevant theory (whenever possible) was discussed to justify and support our conclusions.
In order to successfully model Long Distance Dependencies (LDDs) it is necessary to under-stand the full-range of the characteristics of the LDDs exhibited in a target dataset. In this paper, we use Strictly k-Piecewise languages to generate datasets with various properties. We then compute the characteristics of the LDDs in these datasets using mutual information and analyze the impact of factors such as (i) k, (ii) length of LDDs, (iii) vocabulary size, (iv) forbidden strings, and (v) dataset size. This analysis reveal that the number of interacting elements in a dependency is an important characteristic of LDDs. This leads us to the challenge of modelling multi-element long-distance dependencies. Our results suggest that attention mechanisms in neural networks may aide in modeling datasets with multi-element long-distance dependencies. However, we conclude that there is a need to develop more efficient attention mechanisms to address this issue.
In this paper, we systematically assess the ability of standard recurrent networks to perform dynamic counting and to encode hierarchical representations. All the neural models in our experiments are designed to be small-sized networks both to prevent them from memorizing the training sets and to visualize and interpret their behaviour at test time. Our results demonstrate that the Long Short-Term Memory (LSTM) networks can learn to recognize the well-balanced parenthesis language (Dyck-1) and the shuffles of multiple Dyck-1 languages, each defined over different parenthesis-pairs, by emulating simple real-time k-counter machines. To the best of our knowledge, this work is the first study to introduce the shuffle languages to analyze the computational power of neural networks. We also show that a single-layer LSTM with only one hidden unit is practically sufficient for recognizing the Dyck-1 language. However, none of our recurrent networks was able to yield a good performance on the Dyck-2 language learning task, which requires a model to have a stack-like mechanism for recognition.
Progress in Machine Learning is often driven by the availability of large datasets, and consistent evaluation metrics for comparing modeling approaches. To this end, we present a repository of conversational datasets consisting of hundreds of millions of examples, and a standardised evaluation procedure for conversational response selection models using 1-of-100 accuracy. The repository contains scripts that allow researchers to reproduce the standard datasets, or to adapt the pre-processing and data filtering steps to their needs. We introduce and evaluate several competitive baselines for conversational response selection, whose implementations are shared in the repository, as well as a neural encoder model that is trained on the entire training set.
Conversational machine comprehension (CMC) requires understanding the context of multi-turn dialogue. Using BERT, a pretraining language model, has been successful for single-turn machine comprehension, while modeling multiple turns of question answering with BERT has not been established because BERT has a limit on the number and the length of input sequences. In this paper, we propose a simple but effective method with BERT for CMC. Our method uses BERT to encode a paragraph independently conditioned with each question and each answer in a multi-turn context. Then, the method predicts an answer on the basis of the paragraph representations encoded with BERT. The experiments with representative CMC datasets, QuAC and CoQA, show that our method outperformed recently published methods (+0.8 F1 on QuAC and +2.1 F1 on CoQA). In addition, we conducted a detailed analysis of the effects of the number and types of dialogue history on the accuracy of CMC, and we found that the gold answer history, which may not be given in an actual conversation, contributed to the model performance most on both datasets.
Sequence-to-Sequence (Seq2Seq) models have witnessed a notable success in generating natural conversational exchanges. Notwithstanding the syntactically well-formed responses generated by these neural network models, they are prone to be acontextual, short and generic. In this work, we introduce a Topical Hierarchical Recurrent Encoder Decoder (THRED), a novel, fully data-driven, multi-turn response generation system intended to produce contextual and topic-aware responses. Our model is built upon the basic Seq2Seq model by augmenting it with a hierarchical joint attention mechanism that incorporates topical concepts and previous interactions into the response generation. To train our model, we provide a clean and high-quality conversational dataset mined from Reddit comments. We evaluate THRED on two novel automated metrics, dubbed Semantic Similarity and Response Echo Index, as well as with human evaluation. Our experiments demonstrate that the proposed model is able to generate more diverse and contextually relevant responses compared to the strong baselines.
Response suggestion is an important task for building human-computer conversation systems. Recent approaches to conversation modeling have introduced new model architectures with impressive results, but relatively little attention has been paid to whether these models would be practical in a production setting. In this paper, we describe the unique challenges of building a production retrieval-based conversation system, which selects outputs from a whitelist of candidate responses. To address these challenges, we propose a dual encoder architecture which performs rapid inference and scales well with the size of the whitelist. We also introduce and compare two methods for generating whitelists, and we carry out a comprehensive analysis of the model and whitelists. Experimental results on a large, proprietary help desk chat dataset, including both offline metrics and a human evaluation, indicate production-quality performance and illustrate key lessons about conversation modeling in practice.
This theoretical paper identifies a need for a definition of asymmetric co-creativity where creativity is expected from the computational agent but not from the human user. Our co-operative creativity framework takes into account that the computational agent has a message to convey in a co-operative fashion, which introduces a trade-off on how creative the computer can be. The requirements of co-operation are identified from an interdisciplinary point of view. We divide co-operative creativity in message creativity, contextual creativity and communicative creativity. Finally these notions are applied in the context of the Peace Machine system concept.
We propose a novel method for selecting coherent and diverse responses for a given dialogue context. The proposed method re-ranks response candidates generated from conversational models by using event causality relations between events in a dialogue history and response candidates (e.g., “be stressed out” precedes “relieve stress”). We use distributed event representation based on the Role Factored Tensor Model for a robust matching of event causality relations due to limited event causality knowledge of the system. Experimental results showed that the proposed method improved coherency and dialogue continuity of system responses.
Goal-oriented dialogue in complex domains is an extremely challenging problem and there are relatively few datasets. This task provided two new resources that presented different challenges: one was focused but small, while the other was large but diverse. We also considered several new variations on the next utterance selection problem: (1) increasing the number of candidates, (2) including paraphrases, and (3) not including a correct option in the candidate set. Twenty teams participated, developing a range of neural network models, including some that successfully incorporated external data to boost performance. Both datasets have been publicly released, enabling future work to build on these results, working towards robust goal-oriented dialogue systems.
We tackle the problem of context reconstruction in Chinese dialogue, where the task is to replace pronouns, zero pronouns, and other referring expressions with their referent nouns so that sentences can be processed in isolation without context. Following a standard decomposition of the context reconstruction task into referring expression detection and coreference resolution, we propose a novel end-to-end architecture for separately and jointly accomplishing this task. Key features of this model include POS and position encoding using CNNs and a novel pronoun masking mechanism. One perennial problem in building such models is the paucity of training data, which we address by augmenting previously-proposed methods to generate a large amount of realistic training data. The combination of more data and better models yields accuracy higher than the state-of-the-art method in coreference resolution and end-to-end context reconstruction.
The uncertainties of language and the complexity of dialogue contexts make accurate dialogue state tracking one of the more challenging aspects of dialogue processing. To improve state tracking quality, we argue that relationships between different aspects of dialogue state must be taken into account as they can often guide a more accurate interpretation process. To this end, we present an energy-based approach to dialogue state tracking as a structured classification task. The novelty of our approach lies in the use of an energy network on top of a deep learning architecture to explore more signal correlations between network variables including input features and output labels. We demonstrate that the energy-based approach improves the performance of a deep learning dialogue state tracker towards state-of-the-art results without the need for many of the other steps required by current state-of-the-art methods.
We describe and validate a metric for estimating multi-class classifier performance based on cross-validation and adapted for improvement of small, unbalanced natural-language datasets used in chatbot design. Our experiences draw upon building recruitment chatbots that mediate communication between job-seekers and recruiters by exposing the ML/NLP dataset to the recruiting team. Evaluation approaches must be understandable to various stakeholders, and useful for improving chatbot performance. The metric, nex-cv, uses negative examples in the evaluation of text classification, and fulfils three requirements. First, it is actionable: it can be used by non-developer staff. Second, it is not overly optimistic compared to human ratings, making it a fast method for comparing classifiers. Third, it allows model-agnostic comparison, making it useful for comparing systems despite implementation differences. We validate the metric based on seven recruitment-domain datasets in English and German over the course of one year.
Tracking the state of the conversation is a central component in task-oriented spoken dialogue systems. One such approach for tracking the dialogue state is slot carryover, where a model makes a binary decision if a slot from the context is relevant to the current turn. Previous work on the slot carryover task used models that made independent decisions for each slot. A close analysis of the results show that this approach results in poor performance over longer context dialogues. In this paper, we propose to jointly model the slots. We propose two neural network architectures, one based on pointer networks that incorporate slot ordering information, and the other based on transformer networks that uses self attention mechanism to model the slot interdependencies. Our experiments on an internal dialogue benchmark dataset and on the public DSTC2 dataset demonstrate that our proposed models are able to resolve longer distance slot references and are able to achieve competitive performance.
Dialogue systems and conversational agents are becoming increasingly popular in modern society. We conceptualized one such conversational agent, Microsoft’s “Ruuh” with the promise to be able to talk to its users on any subject they choose. Building an open-ended conversational agent like Ruuh at onset seems like a daunting task, since the agent needs to think beyond the utilitarian notion of merely generating “relevant” responses and meet a wider range of user social needs, like expressing happiness when user’s favourite sports team wins, sharing a cute comment on showing the pictures of the user’s pet and so on. The agent also needs to detect and respond to abusive language, sensitive topics and trolling behaviour of the users. Many of these problems pose significant research challenges as well as product design limitations as one needs to circumnavigate the technical limitations to create an acceptable user experience. However, as the product reaches the real users the true test begins, and one realizes the challenges and opportunities that lie in the vast domain of conversations. With over 2.5 million real-world users till date who have generated over 300 million user conversations with Ruuh, there is a plethora of learning, insights and opportunities that we will talk about in this paper.
Providing plausible responses to why questions is a challenging but critical goal for language based human-machine interaction. Explanations are challenging in that they require many different forms of abstract knowledge and reasoning. Previous work has either relied on human-curated structured knowledge bases or detailed domain representation to generate satisfactory explanations. They are also often limited to ranking pre-existing explanation choices. In our work, we contribute to the under-explored area of generating natural language explanations for general phenomena. We automatically collect large datasets of explanation-phenomenon pairs which allow us to train sequence-to-sequence models to generate natural language explanations. We compare different training strategies and evaluate their performance using both automatic scores and human ratings. We demonstrate that our strategy is sufficient to generate highly plausible explanations for general open-domain phenomena compared to other models trained on different datasets.
We propose an adversarial learning approach for generating multi-turn dialogue responses. Our proposed framework, hredGAN, is based on conditional generative adversarial networks (GANs). The GAN’s generator is a modified hierarchical recurrent encoder-decoder network (HRED) and the discriminator is a word-level bidirectional RNN that shares context and word embeddings with the generator. During inference, noise samples conditioned on the dialogue history are used to perturb the generator’s latent space to generate several possible responses. The final response is the one ranked best by the discriminator. The hredGAN shows improved performance over existing methods: (1) it generalizes better than networks trained using only the log-likelihood criterion, and (2) it generates longer, more informative and more diverse responses with high utterance and topic relevance even with limited training data. This performance improvement is demonstrated on the Movie triples and Ubuntu dialogue datasets with both the automatic and human evaluations.
A sequence-to-sequence model tends to generate generic responses with little information for input utterances. To solve this problem, we propose a neural model that generates relevant and informative responses. Our model has simple architecture to enable easy application to existing neural dialogue models. Specifically, using positive pointwise mutual information, it first identifies keywords that frequently co-occur in responses given an utterance. Then, the model encourages the decoder to use the keywords for response generation. Experiment results demonstrate that our model successfully diversifies responses relative to previous models.
A neural conversation model is a promising approach to develop dialogue systems with the ability of chit-chat. It allows training a model in an end-to-end manner without complex rule design nor feature engineering. However, as a side effect, the neural model tends to generate safe but uninformative and insensitive responses like “OK” and “I don’t know.” Such replies are called generic responses and regarded as a critical problem for user-engagement of dialogue systems. For a more engaging chit-chat experience, we propose a neural conversation model that generates responsive and self-expressive replies. Specifically, our model generates domain-aware and sentiment-rich responses. Experiments empirically confirmed that our model outperformed the sequence-to-sequence model; 68.1% of our responses were domain-aware with sentiment polarities, which was only 2.7% for responses generated by the sequence-to-sequence model.
This paper describes the AX Semantics’ submission to the SIGMORPHON 2019 shared task on morphological reinflection. We implemented two systems, both tackling the task for all languages in one codebase, without any underlying language specific features. The first one is an encoder-decoder model using AllenNLP; the second system uses the same model modified by a custom trainer that trains only with the target language resources after a specific threshold. We especially focused on building an implementation using AllenNLP with out-of-the-box methods to facilitate easy operation and reuse.
We propose cognate projection as a method of crosslingual transfer for inflection generation in the context of the SIGMORPHON 2019 Shared Task. The results on four language pairs show the method is effective when no low-resource training data is available.
We present our CHARLES-SAARLAND system for the SIGMORPHON 2019 Shared Task on Crosslinguality and Context in Morphology, in task 2, Morphological Analysis and Lemmatization in Context. We leverage the multilingual BERT model and apply several fine-tuning strategies introduced by UDify demonstrating exceptional evaluation performance on morpho-syntactic tasks. Our results show that fine-tuning multilingual BERT on the concatenation of all available treebanks allows the model to learn cross-lingual information that is able to boost lemmatization and morphology tagging accuracy over fine-tuning it purely monolingually. Unlike UDify, however, we show that when paired with additional character-level and word-level LSTM layers, a second stage of fine-tuning on each treebank individually can improve evaluation even further. Out of all submissions for this shared task, our system achieves the highest average accuracy and f1 score in morphology tagging and places second in average lemmatization accuracy.
In this paper we describe our system for morphological analysis and lemmatization in context, using a transformer-based sequence to sequence model and a biaffine attention based BiLSTM model. First, a lemma is produced for a given word, and then both the lemma and the given word are used for morphological analysis. We also make use of character level word encodings and trainable encodings to improve accuracy. Overall, our system ranked fifth in lemmatization and sixth in morphological accuracy among twelve systems, and demonstrated considerable improvements over the baseline in morphological analysis.
In this study, we present Morpheus, a joint contextual lemmatizer and morphological tagger. Morpheus is based on a neural sequential architecture where inputs are the characters of the surface words in a sentence and the outputs are the minimum edit operations between surface words and their lemmata as well as the morphological tags assigned to the words. The experiments on the datasets in nearly 100 languages provided by SigMorphon 2019 Shared Task 2 organizers show that the performance of Morpheus is comparable to the state-of-the-art system in terms of lemmatization. In morphological tagging, on the other hand, Morpheus significantly outperforms the SigMorphon baseline. In our experiments, we also show that the neural encoder-decoder architecture trained to predict the minimum edit operations can produce considerably better results than the architecture trained to predict the characters in lemmata directly as in previous studies. According to the SigMorphon 2019 Shared Task 2 results, Morpheus has placed 3rd in lemmatization and reached the 9th place in morphological tagging among all participant teams.
This paper describes our submission to SIGMORPHON 2019 Task 2: Morphological analysis and lemmatization in context. Our model is a multi-task sequence to sequence neural network, which jointly learns morphological tagging and lemmatization. On the encoding side, we exploit character-level as well as contextual information. We introduce a multi-attention decoder to selectively focus on different parts of character and word sequences. To further improve the model, we train on multiple datasets simultaneously and use external embeddings for initialization. Our final model reaches an average morphological tagging F1 score of 94.54 and a lemma accuracy of 93.91 on the test data, ranking respectively 3rd and 6th out of 13 teams in the SIGMORPHON 2019 shared task.
This paper presents the Instituto de Telecomunicações–Instituto Superior Técnico submission to Task 1 of the SIGMORPHON 2019 Shared Task. Our models combine sparse sequence-to-sequence models with a two-headed attention mechanism that learns separate attention distributions for the lemma and inflectional tags. Among submissions to Task 1, our models rank second and third. Despite the low data setting of the task (only 100 in-language training examples), they learn plausible inflection patterns and often concentrate all probability mass into a small set of hypotheses, making beam search exact.
This paper presents the submission by the CMU-01 team to the SIGMORPHON 2019 task 2 of Morphological Analysis and Lemmatization in Context. This task requires us to produce the lemma and morpho-syntactic description of each token in a sequence, for 107 treebanks. We approach this task with a hierarchical neural conditional random field (CRF) model which predicts each coarse-grained feature (eg. POS, Case, etc.) independently. However, most treebanks are under-resourced, thus making it challenging to train deep neural models for them. Hence, we propose a multi-lingual transfer training regime where we transfer from multiple related languages that share similar typology.
This paper describes two related systems for cross-lingual morphological inflection for SIGMORPHON 2019 Shared Task participation. Both sets of results submitted to the shared task for evaluation are obtained using a simple approach of predicting transducer actions based on initial alignments on the training set, where cross-lingual transfer is limited to only using the high-resource language data as additional training set. The performance of the system does not reach the performance of the top two systems in the competition. However, we show that results can be improved with further tuning. We also present further analyses demonstrating that the cross-lingual gain is rather modest.
This paper describes the OSU submission to the SIGMORPHON 2019 shared task, Crosslinguality and Context in Morphology. Our system addresses the contextual morphological analysis subtask of Task 2, which is to produce the morphosyntactic description (MSD) of each fully inflected word within a given sentence. We frame this as a sequence generation task and employ a neural encoder-decoder (seq2seq) architecture to generate the sequence of MSD tags given the encoded representation of each token. Follow-up analyses reveal that our system most significantly improves performance on morphologically complex languages whose inflected word forms typically have longer MSD tag sequences. In addition, our system seems to capture the structured correlation between MSD tags, such as that between the “verb” tag and TAM-related tags.
This paper presents the UNT HiLT+Ling system for the Sigmorphon 2019 shared Task 2: Morphological Analysis and Lemmatization in Context. Our core approach focuses on the morphological tagging task; part-of-speech tagging and lemmatization are treated as secondary tasks. Given the highly multilingual nature of the task, we propose an approach which makes minimal use of the supplied training data, in order to be extensible to languages without labeled training data for the morphological inflection task. Specifically, we use a parallel Bible corpus to align contextual embeddings at the verse level. The aligned verses are used to build cross-language translation matrices, which in turn are used to map between embedding spaces for the various languages. Finally, we use sets of inflected forms, primarily from a high-resource language, to induce vector representations for individual UniMorph tags. Morphological analysis is performed by matching vector representations to embeddings for individual tokens. While our system results are dramatically below the average system submitted for the shared task evaluation campaign, our method is (we suspect) unique in its minimal reliance on labeled training data.
We present our contribution to the SIGMORPHON 2019 Shared Task: Crosslinguality and Context in Morphology, Task 2: contextual morphological analysis and lemmatization. We submitted a modification of the UDPipe 2.0, one of best-performing systems of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies and an overall winner of the The 2018 Shared Task on Extrinsic Parser Evaluation. As our first improvement, we use the pretrained contextualized embeddings (BERT) as additional inputs to the network; secondly, we use individual morphological features as regularization; and finally, we merge the selected corpora of the same language. In the lemmatization task, our system exceeds all the submitted systems by a wide margin with lemmatization accuracy 95.78 (second best was 95.00, third 94.46). In the morphological analysis, our system placed tightly second: our morphological analysis accuracy was 93.19, the winning system’s 93.23.
This paper presents the submission by the Charles University-University of Malta team to the SIGMORPHON 2019 Shared Task on Morphological Analysis and Lemmatization in context. We present a lemmatization model based on previous work on neural transducers (Makarov and Clematide, 2018b; Aharoni and Goldberg, 2016). The key difference is that our model transforms the whole word form in every step, instead of consuming it character by character. We propose a merging strategy inspired by Byte-Pair-Encoding that reduces the space of valid operations by merging frequent adjacent operations. The resulting operations not only encode the actions to be performed but the relative position in the word token and how characters need to be transformed. Our morphological tagger is a vanilla biLSTM tagger that operates over operation representations, encoding operations and words in a hierarchical manner. Even though relative performance according to metrics is below the baseline, experiments show that our models capture important associations between interpretable operation labels and fine-grained morpho-syntax labels.
We present de-lexical segmentation, a linguistically motivated alternative to greedy or other unsupervised methods, requiring only minimal language specific input. Our technique involves creating a small grammar of closed-class affixes which can be written in a few hours. The grammar over generates analyses for word forms attested in a raw corpus which are disambiguated based on features of the linguistic base proposed for each form. Extending the grammar to cover orthographic, morpho-syntactic or lexical variation is simple, making it an ideal solution for challenging corpora with noisy, dialect-inconsistent, or otherwise non-standard content. In two evaluations, we consistently outperform competitive unsupervised baselines and approach the performance of state-of-the-art supervised models trained on large amounts of data, providing evidence for the value of linguistic input during preprocessing.
We show that MaxEnt is so rich that it can distinguish between any two different mappings: there always exists a nonnegative weight vector which assigns them different MaxEnt probabilities. Stochastic HG instead does admit equiprobable mappings and we give a complete formal characterization of them.
This paper situates culminative unbounded stress systems within the subregular hierarchy for functions. While Baek (2018) has argued that such systems can be uniformly understood as input tier-based strictly local constraints, we show here that default-to-opposite-side and default-to-same-side stress systems belong to distinct subregular classes when they are viewed as functions that assign primary stress to underlying forms. While the former system can be captured by input tier-based input strictly local functions, a subsequential function class that we define here, the latter system is not subsequential, though it is weakly deterministic according to McCollum et al.’s (2018) non-interaction criterion. Our results motivate the extension of recently proposed subregular language classes to subregular functions and argue in favor of McCollum et al’s definition of weak determinism over that of Heinz and Lai (2013).
In spontaneous speech, Mandarin tones that belong to the same tone category may exhibit many different contour shapes. We explore the use of time-series data mining techniques for understanding the variability of tones in a large corpus of Mandarin newscast speech. First, we adapt a graph-based approach to characterize the clusters (fuzzy types) of tone contour shapes observed in each tone n-gram category. Second, we show correlations between these realized contour shape clusters and a bag of automatically extracted linguistic features. We discuss the implications of the current study within the context of phonological and information theory.
We apply convolutional neural networks to the task of shallow morpheme segmentation using low-resource datasets for 5 different languages. We show that both in fully supervised and semi-supervised settings our model beats previous state-of-the-art approaches. We argue that convolutional neural networks reflect local nature of morpheme segmentation better than other semi-supervised approaches.
Recent work has looked at evaluation of phone embeddings using sound analogies and correlations between distinctive feature space and embedding space. It has not been clear what aspects of natural language phonology are learnt by neural network inspired distributed representational models such as word2vec. To study the kinds of phonological relationships learnt by phone embeddings, we present artificial phonology experiments that show that phone embeddings learn paradigmatic relationships such as phonemic and allophonic distribution quite well. They are also able to capture co-occurrence restrictions among vowels such as those observed in languages with vowel harmony. However, they are unable to learn co-occurrence restrictions among the class of consonants.
Previous “wug” tests (Berko, 1958) on Japanese verbal inflection have demonstrated that Japanese speakers, both adults and children, cannot inflect novel present tense forms to “correct” past tense forms predicted by rules of existent verbs (de Chene, 1982; Vance, 1987, 1991; Klafehn, 2003, 2013), indicating that Japanese verbs are merely stored in the mental lexicon. However, the implicit assumption that present tense forms are bases for verbal inflection should not be blindly extended to morphologically rich languages like Japanese in which both present and past tense forms are morphologically complex without inherent direction (Albright, 2002). Interestingly, there are also independent observations in the acquisition literature to suggest that past tense forms may be bases for verbal inflection in Japanese (Klafehn, 2003; Murasugi et al., 2010; Hirose, 2017; Tatsumi et al., 2018). In this paper, we computationally simulate two directions of verbal inflection in Japanese, Present → Past and Past → Present, with the rule-based computational model called Minimal Generalization Learner (MGL; Albright and Hayes, 2003) and experimentally evaluate the model with the bidirectional “wug” test where humans inflect novel verbs in two opposite directions. We conclude that Japanese verbs can be computed online via some generalizations and those generalizations do depend on the direction of morphological inflection.
This paper deals with the automatic enhancement of a new German morphological database. While there are some databases for flat word segmentation, this is the first available resource which can be directly used for deep parsing of German words. We combine the entries of this morphological database with the morphological tools SMOR and Moremorph and a context-based evaluation method which builds on a large Wikipedia corpus. We describe the state of the art and the essential characteristics of the database and the context method. The approach is tested on an inflight magazine of Lufthansa. We derive over 5,000 new instances of complex words. The coverage for the lemma types reaches up to over 99 percent. The precision of new found complex splits and monomorphemes is between 0.93 and 0.99.
Polysynthetic languages pose a challenge for morphological analysis due to the root-morpheme complexity and to the word class “squish”. In addition, many of these polysynthetic languages are low-resource. We propose unsupervised approaches for morphological segmentation of low-resource polysynthetic languages based on Adaptor Grammars (AG) (Eskander et al., 2016). We experiment with four languages from the Uto-Aztecan family. Our AG-based approaches outperform other unsupervised approaches and show promise when compared to supervised methods, outperforming them on two of the four languages.
Whether phonological transformations in general are subregular is an open question. This is the case for most transformations, which have been shown to be subsequential, but it is not known whether weakly deterministic mappings form a proper subset of the regular functions. This paper demonstrates that there are regular functions that are not weakly deterministic, and, because all attested processes are weakly deterministic, supports the subregular hypothesis.
We use sequence-to-sequence networks trained on sequential phonetic encoding tasks to construct compositional phonological representations of words. We show that the output of an encoder network can predict the phonetic durations of American English words better than a number of alternative forms. We also show that the model’s learned representations map onto existing measures of words’ phonological structure (phonological neighborhood density and phonotactic probability).
This paper defines a subregular class of functions called the tier-based synchronized strictly local (TSSL) functions. These functions are similar to the the tier-based input-output strictly local (TIOSL) functions, except that the locality condition is enforced not on the input and output streams, but on the computation history of the minimal subsequential finite-state transducer. We show that TSSL functions naturally describe rhythmic syncope while TIOSL functions cannot, and we argue that TSSL functions provide a more restricted characterization of rhythmic syncope than existing treatments within Optimality Theory.
The SIGMORPHON 2019 shared task on cross-lingual transfer and contextual analysis in morphology examined transfer learning of inflection between 100 language pairs, as well as contextual lemmatization and morphosyntactic description in 66 languages. The first task evolves past years’ inflection tasks by examining transfer of morphological inflection knowledge from a high-resource language to a low-resource language. This year also presents a new second challenge on lemmatization and morphological feature analysis in context. All submissions featured a neural component and built on either this year’s strong baselines or highly ranked systems from previous years’ shared tasks. Every participating team improved in accuracy over the baselines for the inflection task (though not Levenshtein distance), and every team in the contextual analysis task improved on both state-of-the-art neural and non-neural baselines.
The presence of toxic content has become a major problem for many online communities. Moderators try to limit this problem by implementing more and more refined comment filters, but toxic users are constantly finding new ways to circumvent them. Our hypothesis is that while modifying toxic content and keywords to fool filters can be easy, hiding sentiment is harder. In this paper, we explore various aspects of sentiment detection and their correlation to toxicity, and use our results to implement a toxicity detection tool. We then test how adding the sentiment information helps detect toxicity in three different real-world datasets, and incorporate subversion to these datasets to simulate a user trying to circumvent the system. Our results show sentiment information has a positive impact on toxicity detection.
Interactions among users on social network platforms are usually positive, constructive and insightful. However, sometimes people also get exposed to objectionable content such as hate speech, bullying, and verbal abuse etc. Most social platforms have explicit policy against hate speech because it creates an environment of intimidation and exclusion, and in some cases may promote real-world violence. As users’ interactions on today’s social networks involve multiple modalities, such as texts, images and videos, in this paper we explore the challenge of automatically identifying hate speech with deep multimodal technologies, extending previous research which mostly focuses on the text signal alone. We present a number of fusion approaches to integrate text and photo signals. We show that augmenting text with image embedding information immediately leads to a boost in performance, while applying additional attention fusion methods brings further improvement.
We developed a machine-learning-based method to detect video game players that harass teammates or opponents in chat earlier in the conversation. This real-time technology would allow gaming companies to intervene during games, such as issue warnings or muting or banning a player. In a proof-of-concept experiment on League of Legends data we compute and visualize evaluation metrics for a machine learning classifier as conversations unfold, and observe that the optimal precision and recall of detecting toxic players at each moment in the conversation depends on the confidence threshold of the classifier: the threshold should start low, and increase as the conversation unfolds. How fast this sliding threshold should increase depends on the training set size.
Technologies for abusive language detection are being developed and applied with little consideration of their potential biases. We examine racial bias in five different sets of Twitter data annotated for hate speech and abusive language. We train classifiers on these datasets and compare the predictions of these classifiers on tweets written in African-American English with those written in Standard American English. The results show evidence of systematic racial bias in all datasets, as classifiers trained on them tend to predict that tweets written in African-American English are abusive at substantially higher rates. If these abusive language detection systems are used in the field they will therefore have a disproportionate negative impact on African-American social media users. Consequently, these systems may discriminate against the groups who are often the targets of the abuse we are trying to detect.
Discussion forum participation represents one of the crucial factors for learning and often the only way of supporting social interactions in online settings. However, as much as sharing new ideas or asking thoughtful questions contributes learning, verbally abusive behaviors, such as expressing negative emotions in online discussions, could have disproportionate detrimental effects. To provide means for mitigating the potential negative effects on course participation and learning, we developed an automated classifier for identifying communication that show linguistic patterns associated with hostility in online forums. In so doing, we employ several well-established automated text analysis tools and build on the common practices for handling highly imbalanced datasets and reducing the sensitivity to overfitting. Although still in its infancy, our approach shows promising results (ROC AUC .73) towards establishing a robust detector of abusive behaviors. We further provide an overview of the classification (linguistic and contextual) features most indicative of online aggression.
Hate speech and abusive language spreading on social media need to be detected automatically to avoid conflict between citizen. Moreover, hate speech has a target, category, and level that also needs to be detected to help the authority in prioritizing which hate speech must be addressed immediately. This research discusses multi-label text classification for abusive language and hate speech detection including detecting the target, category, and level of hate speech in Indonesian Twitter using machine learning approach with Support Vector Machine (SVM), Naive Bayes (NB), and Random Forest Decision Tree (RFDT) classifier and Binary Relevance (BR), Label Power-set (LP), and Classifier Chains (CC) as the data transformation method. We used several kinds of feature extractions which are term frequency, orthography, and lexicon features. Our experiment results show that in general RFDT classifier using LP as the transformation method gives the best accuracy with fast computational time.
Recent concerns over abusive behavior on their platforms have pressured social media companies to strengthen their content moderation policies. However, user opinions on these policies have been relatively understudied. In this paper, we present an analysis of user responses to a September 27, 2018 announcement about the quarantine policy on Reddit as a case study of to what extent the discourse on content moderation is polarized by users’ ideological viewpoint. We introduce a novel partitioning approach for characterizing user polarization based on their distribution of participation across interest subreddits. We then use automated techniques for capturing framing to examine how users with different viewpoints discuss moderation issues, finding that right-leaning users invoked censorship while left-leaning users highlighted inconsistencies on how content policies are applied. Overall, we argue for a more nuanced approach to moderation by highlighting the intersection of behavior and ideology in considering how abusive language is defined and regulated.
The goal of any social media platform is to facilitate healthy and meaningful interactions among its users. But more often than not, it has been found that it becomes an avenue for wanton attacks. We propose an experimental study that has three aims: 1) to provide us with a deeper understanding of current data sets that focus on different types of abusive language, which are sometimes overlapping (racism, sexism, hate speech, offensive language, and personal attacks); 2) to investigate what type of attention mechanism (contextual vs. self-attention) is better for abusive language detection using deep learning architectures; and 3) to investigate whether stacked architectures provide an advantage over simple architectures for this task.
Online abusive content detection is an inherently difficult task. It has received considerable attention from academia, particularly within the computational linguistics community, and performance appears to have improved as the field has matured. However, considerable challenges and unaddressed frontiers remain, spanning technical, social and ethical dimensions. These issues constrain the performance, efficiency and generalizability of abusive content detection systems. In this article we delineate and clarify the main challenges and frontiers in the field, critically evaluate their implications and discuss potential solutions. We also highlight ways in which social scientific insights can advance research. We discuss the lack of support given to researchers working with abusive content and provide guidelines for ethical research.
Over the past years, the amount of online offensive speech has been growing steadily. To successfully cope with it, machine learning are applied. However, ML-based techniques require sufficiently large annotated datasets. In the last years, different datasets were published, mainly for English. In this paper, we present a new dataset for Portuguese, which has not been in focus so far. The dataset is composed of 5,668 tweets. For its annotation, we defined two different schemes used by annotators with different levels of expertise. Firstly, non-experts annotated the tweets with binary labels (‘hate’ vs. ‘no-hate’). Secondly, expert annotators classified the tweets following a fine-grained hierarchical multiple label scheme with 81 hate speech categories in total. The inter-annotator agreement varied from category to category, which reflects the insight that some types of hate speech are more subtle than others and that their detection depends on personal perception. This hierarchical annotation scheme is the main contribution of the presented work, as it facilitates the identification of different types of hate speech and their intersections. To demonstrate the usefulness of our dataset, we carried a baseline classification experiment with pre-trained word embeddings and LSTM on the binary classified data, with a state-of-the-art outcome.
Social media platforms like Twitter and Instagram face a surge in cyberbullying phenomena against young users and need to develop scalable computational methods to limit the negative consequences of this kind of abuse. Despite the number of approaches recently proposed in the Natural Language Processing (NLP) research area for detecting different forms of abusive language, the issue of identifying cyberbullying phenomena at scale is still an unsolved problem. This is because of the need to couple abusive language detection on textual message with network analysis, so that repeated attacks against the same person can be identified. In this paper, we present a system to monitor cyberbullying phenomena by combining message classification and social network analysis. We evaluate the classification module on a data set built on Instagram messages, and we describe the cyberbullying monitoring user interface.
Hate speech and abusive language have become a common phenomenon on Arabic social media. Automatic hate speech and abusive detection systems can facilitate the prohibition of toxic textual contents. The complexity, informality and ambiguity of the Arabic dialects hindered the provision of the needed resources for Arabic abusive/hate speech detection research. In this paper, we introduce the first publicly-available Levantine Hate Speech and Abusive (L-HSAB) Twitter dataset with the objective to be a benchmark dataset for automatic detection of online Levantine toxic contents. We, further, provide a detailed review of the data collection steps and how we design the annotation guidelines such that a reliable dataset annotation is guaranteed. This has been later emphasized through the comprehensive evaluation of the annotations as the annotation agreement metrics of Cohen’s Kappa (k) and Krippendorff’s alpha (α) indicated the consistency of the annotations.
In this paper, we describe a workflow for the data-driven acquisition and semantic scaling of a lexicon that covers lexical items from the lower end of the German language register—terms typically considered as rough, vulgar or obscene. Since the fine semantic representation of grades of obscenity can only inadequately be captured at the categorical level (e.g., obscene vs. non-obscene, or rough vs. vulgar), our main contribution lies in applying best-worst scaling, a rating methodology that has already been shown to be useful for emotional language, to capture the relative strength of obscenity of lexical items. We describe the empirical foundations for bootstrapping such a low-end lexicon for German by starting from manually supplied lexicographic categorizations of a small seed set of rough and vulgar lexical items and automatically enlarging this set by means of distributional semantics. We then determine the degrees of obscenity for the full set of all acquired lexical items by letting crowdworkers comparatively assess their pejorative grade using best-worst scaling. This semi-automatically enriched lexicon already comprises 3,300 lexical items and incorporates 33,000 vulgarity ratings. Using it as a seed lexicon for fully automatic lexical acquisition, we were able to raise its coverage up to slightly more than 11,000 entries.
We address the task of automatically detecting toxic content in user generated texts. We fo cus on exploring the potential for preemptive moderation, i.e., predicting whether a particular conversation thread will, in the future, incite a toxic comment. Moreover, we perform preliminary investigation of whether a model that jointly considers all comments in a conversation thread outperforms a model that considers only individual comments. Using an existing dataset of conversations among Wikipedia contributors as a starting point, we compile a new large-scale dataset for this task consisting of labeled comments and comments from their conversation threads.
The text we see in social media suffers from lots of undesired characterstics like hatespeech, abusive language, insults etc. The nature of this text is also very different compared to the traditional text we see in news with lots of obfuscated words, intended typos. This poses several robustness challenges to many natural language processing (NLP) techniques developed for traditional text. Many techniques proposed in the recent times such as charecter encoding models, subword models, byte pair encoding to extract subwords can aid in dealing with few of these nuances. In our work, we analyze the effectiveness of each of the above techniques, compare and contrast various word decomposition techniques when used in combination with others. We experiment with recent advances of finetuning pretrained language models, and demonstrate their robustness to domain shift. We also show our approaches achieve state of the art performance on Wikipedia attack, toxicity datasets, and Twitter hatespeech dataset.
Hate speech detectors must be applicable across a multitude of services and platforms, and there is hence a need for detection approaches that do not depend on any information specific to a given platform. For instance, the information stored about the text’s author may differ between services, and so using such data would reduce a system’s general applicability. The paper thus focuses on using exclusively text-based input in the detection, in an optimised architecture combining Convolutional Neural Networks and Long Short-Term Memory-networks. The hate speech detector merges two strands with character n-grams and word embeddings to produce the final classification, and is shown to outperform comparable previous approaches.
In the era of social media, hate speech, trolling and verbal abuse have become a common issue. We present an approach to automatically classify such statements, using a new deep learning architecture. Our model comprises of a Multi Dimension Capsule Network that generates the representation of sentences which we use for classification. We further provide an analysis of our model’s interpretation of such statements. We compare the results of our model with state-of-art classification algorithms and demonstrate our model’s ability. It also has the capability to handle comments that are written in both Hindi and English, which are provided in the TRAC dataset. We also compare results on Kaggle’s Toxic comment classification dataset.
The paper proposes an investigation on the role of populist themes and rhetoric in an Italian Twitter corpus of hate speech against immigrants. The corpus had been annotated with four new layers of analysis: Nominal Utterances, that can be seen as consistent with populist rhetoric; In-out-group rhetoric, a very common populist strategy to polarize public opinion; Slogan-like nominal utterances, that may convey the call for severe illiberal policies against immigrants; News, to recognize the role of newspapers (headlines or reference to articles) in the Twitter political discourse on immigration featured by hate speech.
The disciplines of Gender Studies and Data Science are incompatible. This is conventional wisdom, supported by how many computational studies simplify gender into an immutable binary categorization that appears crude to the critical social researcher. I argue that the characterization of gender norms is context specific and may prove valuable in constructing useful models. I show how gender can be framed in computational studies as a stylized repetition of acts mediated by a social structure, and not a possessed biological category. By conducting a review of existing work, I show how gender should be explored in multiplicity in computational research through clustering techniques, and layout how this is being achieved in a study in progress on gender hostility on Stack Overflow.
The present paper introduces a theoretical model for explaining aggressive online comments from a sociological perspective. It is innovative as it combines individual, situational, and social-structural determinants of online aggression and tries to theoretically derive their interplay. Moreover, the paper suggests an empirical strategy for testing the model. The main contribution will be to match online commenting data with survey data containing rich background data of non- /aggressive online commentators.
The segmentation of argumentative units is an important subtask of argument mining, which is frequently addressed at a coarse granularity, usually assuming argumentative units to be no smaller than sentences. Approaches focusing at the clause-level granularity, typically address the task as sequence labeling at the token level, aiming to classify whether a token begins, is inside, or is outside of an argumentative unit. Most approaches exploit highly engineered, manually constructed features, and algorithms typically used in sequential tagging – such as Conditional Random Fields, while more recent approaches try to exploit manually constructed features in the context of deep neural networks. In this context, we examined to what extend recent advances in sequential labelling allow to reduce the need for highly sophisticated, manually constructed features, and whether limiting features to embeddings, pre-trained on large corpora is a promising approach. Evaluation results suggest the examined models and approaches can exhibit comparable performance, minimising the need for feature engineering.
We present a model to tackle a fundamental but understudied problem in computational argumentation: proposition extraction. Propositions are the basic units of an argument and the primary building blocks of most argument mining systems. However, they are usually substituted by argumentative discourse units obtained via surface-level text segmentation, which may yield text segments that lack semantic information necessary for subsequent argument mining processes. In contrast, our cascade model aims to extract complete propositions by handling anaphora resolution, text segmentation, reported speech, questions, imperatives, missing subjects, and revision. We formulate each task as a computational problem and test various models using a corpus of the 2016 U.S. presidential debates. We show promising performance for some tasks and discuss main challenges in proposition extraction.
When assessing relations between argumentative units (e.g., support or attack), computational systems often exploit disclosing indicators or markers that are not part of elementary argumentative units (EAUs) themselves, but are gained from their context (position in paragraph, preceding tokens, etc.). We show that this dependency is much stronger than previously assumed. In fact, we show that by completely masking the EAU text spans and only feeding information from their context, a competitive system may function even better. We argue that an argument analysis system that relies more on discourse context than the argument’s content is unsafe, since it can easily be tricked. To alleviate this issue, we separate argumentative units from their context such that the system is forced to model and rely on an EAU’s content. We show that the resulting classification system is more robust, and argue that such models are better suited for predicting argumentative relations across documents.
In this paper, we investigate similarities between discourse and argumentation structures by aligning subtrees in a corpus containing both annotations. Contrary to previous works, we focus on comparing sub-structures and not only relations matches. Using data mining techniques, we show that discourse and argumentation most often align well, and the double annotation allows to derive a mapping between structures. Moreover, this approach enables the study of similarities between discourse structures and differences in their expressive power.
In this work we propose to leverage resources available with discourse-level annotations to facilitate the identification of argumentative components and relations in scientific texts, which has been recognized as a particularly challenging task. In particular, we implement and evaluate a transfer learning approach in which contextualized representations learned from discourse parsing tasks are used as input of argument mining models. As a pilot application, we explore the feasibility of using automatically identified argumentative components and relations to predict the acceptance of papers in computer science venues. In order to conduct our experiments, we propose an annotation scheme for argumentative units and relations and use it to enrich an existing corpus with an argumentation layer.
As part of a larger project on argument mining of Swedish parliamentary data, we have created a semantic graph that, together with named entity recognition and resolution (NER), should make it easier to establish connections between arguments in a given debate. The graph is essentially a semantic database that keeps track of Members of Parliament (MPs), in particular their presence in the parliament and activity in debates, but also party affiliation and participation in commissions. The hope is that the Swedish PoliGraph will enable us to perform named entity resolution on debates in the Swedish parliament with a high accuracy, with the aim of determining to whom an argument is directed.
Engaging in a live debate requires, among other things, the ability to effectively rebut arguments claimed by your opponent. In particular, this requires identifying these arguments. Here, we suggest doing so by automatically mining claims from a corpus of news articles containing billions of sentences, and searching for them in a given speech. This raises the question of whether such claims indeed correspond to those made in spoken speeches. To this end, we collected a large dataset of 400 speeches in English discussing 200 controversial topics, mined claims for each topic, and asked annotators to identify the mined claims mentioned in each speech. Results show that in the vast majority of speeches debaters indeed make use of such claims. In addition, we present several baselines for the automatic detection of mined claims in speeches, forming the basis for future work. All collected data is freely available for research.
Identification of argumentative components is an important stage of argument mining. Lexicon information is reported as one of the most frequently used features in the argument mining research. In this paper, we propose a methodology to integrate lexicon information into a neural network model by attention mechanism. We conduct experiments on the UKP dataset, which is collected from heterogeneous sources and contains several text types, e.g., microblog, Wikipedia, and news. We explore lexicons from various application scenarios such as sentiment analysis and emotion detection. We also compare the experimental results of leveraging different lexicons.
Attention mechanisms have seen some success for natural language processing downstream tasks in recent years and generated new state-of-the-art results. A thorough evaluation of the attention mechanism for the task of Argumentation Mining is missing. With this paper, we report a comparative evaluation of attention layers in combination with a bidirectional long short-term memory network, which is the current state-of-the-art approach for the unit segmentation task. We also compare sentence-level contextualized word embeddings to pre-generated ones. Our findings suggest that for this task, the additional attention layer does not improve the performance. In most cases, contextualized embeddings do also not show an improvement on the score achieved by pre-defined embeddings.
In recent years, argumentation mining, which automatically extracts the structure of argumentation from unstructured documents such as essays and debates, is gaining attention. For argumentation mining applications, argument-component classification is an important subtask. The existing methods can be classified into supervised methods and unsupervised methods. Supervised document classification performs classification using a single sentence without relying on the whole document. On the other hand, unsupervised document classification has the advantage of being able to use the whole document, but accuracy of these methods is not so high. In this paper, we propose a method for argument-component classification that combines relation identification by neural networks and TextRank to integrate relation informations (i.e. the strength of the relation). This method can use argumentation-specific knowledge by employing a supervised learning on a corpus while maintaining the advantage of using the whole document. Experiments on two corpora, one consisting of student essay and the other of Wikipedia articles, show the effectiveness of this method.
The purpose of this study is to deploy a novel methodology for classifying different argumentative support (supporting evidences) in arguments, without considering the context. The proposed methodology is based on the idea that the use of Tree Kernel algorithms can be a good way to discriminate between different types of argumentative stances without the need of highly engineered features. This can be useful in different Argumentation Mining sub-tasks. This work provides an example of classifier built using a Tree Kernel method, which can discriminate between different kinds of argumentative support with a high accuracy. The ability to distinguish different kinds of support is, in fact, a key step toward Argument Scheme classification.
Research on argumentation mining from text has frequently discussed relationships to discourse parsing, but few empirical results are available so far. One corpus that has been annotated in parallel for argumentation structure and for discourse structure (RST, SDRT) are the ‘argumentative microtexts’ (Peldszus and Stede, 2016a). While results on perusing the gold RST annotations for predicting argumentation have been published (Peldszus and Stede, 2016b), the step to automatic discourse parsing has not yet been taken. In this paper, we run various discourse parsers (RST, PDTB) on the corpus, compare their results to the gold annotations (for RST) and then assess the contribution of automatically-derived discourse features for argumentation parsing. After reproducing the state-of-the-art Evidence Graph model from Afantenos et al. (2018) for the microtexts, we find that PDTB features can indeed improve its performance.
We report the results of preliminary investigations into the relationship between linguistic alignment and dialogical argumentation at the level of discourse acts. We annotated a proof of concept dataset with illocutions and transitions at the comment level based on Inference Anchoring Theory. We estimated linguistic alignment across discourse acts and found significant variation. Alignment features calculated at the dyad level are found to be useful for detecting a range of argumentative discourse acts.
This paper focuses on the real world application of scientific writing and on determining rhetorical moves, an important step in establishing the argument structure of biomedical articles. Using the observation that the structure of scholarly writing in laboratory-based experimental sciences closely follows laboratory procedures, we examine most closely the Methods section of the texts and adopt an approach of identifying rhetorical moves that are procedure-oriented. We also propose a verb-centric frame semantics with an effective set of semantic roles in order to support the analysis. These components are designed to support a computational model that extends a promising proposal of appropriate rhetorical moves for this domain, but one which is merely descriptive. Our work also contributes to the understanding of argument-related annotation schemes. In particular, we conduct a detailed study with human annotators to confirm that our selection of semantic roles is effective in determining the underlying rhetorical structure of existing biomedical articles in an extensive dataset. The annotated dataset that we produce provides the important knowledge needed for our ultimate goal of analyzing biochemistry articles.
Rhetorical elements from scientific publications provide a more structured view of the document and allow algorithms to focus on particular parts of the text. We surveyed the literature for previously proposed schemes for rhetorical elements and present an overview of its current state of the art. We also searched for available tools using these schemes and applied four tools for our particular task of ranking biomedical abstracts based on text similarity. Comparison of the tools with two strong baselines shows that the predictions provided by the ArguminSci tool can support our use case of mining alternative methods for animal experiments.
We tackle the tasks of automatically identifying comparative sentences and categorizing the intended preference (e.g., “Python has better NLP libraries than MATLAB” → Python, better, MATLAB). To this end, we manually annotate 7,199 sentences for 217 distinct target item pairs from several domains (27% of the sentences contain an oriented comparison in the sense of “better” or “worse”). A gradient boosting model based on pre-trained sentence embeddings reaches an F1 score of 85% in our experimental evaluation. The model can be used to extract comparative sentences for pro/con argumentation in comparative / argument search engines or debating technologies.
In data ranking applications, pairwise annotation is often more consistent than cardinal annotation for learning ranking models. We examine this in a case study on ranking text passages for argument convincingness. Our task is to choose text passages that provide the highest-quality, most-convincing arguments for opposing sides of a topic. Using data from a deployed system within the Bing search engine, we construct a pairwise-labeled dataset for argument convincingness that is substantially more comprehensive in topical coverage compared to existing public resources. We detail the process of extracting topical passages for queries submitted to a search engine, creating annotated sets of passages aligned to different stances on a topic, and assessing argument convincingness of passages using pairwise annotation. Using a state-of-the-art convincingness model, we evaluate several methods for using pairwise-annotated data examples to train models for ranking passages. Our results show pairwise training outperforms training that regresses to a target score for each passage. Our results also show a simple ‘win-rate’ score is a better regression target than the previously proposed page-rank target. Lastly, addressing the need to filter noisy crowd-sourced annotations when constructing a dataset, we show that filtering for transitivity within pairwise annotations is more effective than filtering based on annotation confidence measures for individual examples.
Stance detection plays a pivot role in fake news detection. The task involves determining the point of view or stance – for or against – a text takes towards a claim. One very important stage in employing stance detection for fake news detection is the aggregation of multiple stance labels from different text sources in order to compute a prediction for the veracity of a claim. Typically, aggregation is treated as a credibility-weighted average of stance predictions. In this work, we take the novel approach of applying, for aggregation, a gradual argumentation semantics to bipolar argumentation frameworks mined using stance detection. Our empirical evaluation shows that our method results in more accurate veracity predictions.
This paper examines the factors that govern persuasion for a priori UNDECIDED versus DECIDED audience members in the context of on-line debates. We separately study two types of influences: linguistic factors — features of the language of the debate itself; and audience factors — features of an audience member encoding demographic information, prior beliefs, and debate platform behavior. In a study of users of a popular debate platform, we find first that different combinations of linguistic features are critical for predicting persuasion outcomes for UNDECIDED versus DECIDED members of the audience. We additionally find that audience factors have more influence on predicting the side (PRO/CON) that persuaded UNDECIDED users than for DECIDED users that flip their stance to the opposing side. Our results emphasize the importance of considering the undecided and decided audiences separately when studying linguistic factors of persuasion.
This paper presents a first attempt at using Walton’s argumentation schemes for annotating arguments in Swedish political text and assessing the feasibility of using this particular set of schemes with two linguistically trained annotators. The texts are not pre-annotated with argumentation structure beforehand. The results show that the annotators differ both in number of annotated arguments and selection of the conclusion and premises which make up the arguments. They also differ in their labeling of the schemes, but grouping the schemes increases their agreement. The outcome from this will be used to develop guidelines for future annotations.
The issues of algorithmic fairness and bias have recently featured prominently in many publications highlighting the fact that training the algorithms for maximum performance may often result in predictions that are biased against various groups. Educational applications based on NLP and speech processing technologies often combine multiple complex machine learning algorithms and are thus vulnerable to the same sources of bias as other machine learning systems. Yet such systems can have high impact on people’s lives especially when deployed as part of high-stakes tests. In this paper we discuss different definitions of fairness and possible ways to apply them to educational applications. We then use simulated and real data to consider how test-takers’ native language backgrounds can affect their automated scores on an English language proficiency assessment. We illustrate that total fairness may not be achievable and that different definitions of fairness may require different solutions.
Predicting the construct-relevant difficulty of Multiple-Choice Questions (MCQs) has the potential to reduce cost while maintaining the quality of high-stakes exams. In this paper, we propose a method for estimating the difficulty of MCQs from a high-stakes medical exam, where all questions were deliberately written to a common reading level. To accomplish this, we extract a large number of linguistic features and embedding types, as well as features quantifying the difficulty of the items for an automatic question-answering system. The results show that the proposed approach outperforms various baselines with a statistically significant difference. Best results were achieved when using the full feature set, where embeddings had the highest predictive power, followed by linguistic features. An ablation study of the various types of linguistic features suggested that information from all levels of linguistic processing contributes to predicting item difficulty, with features related to semantic ambiguity and the psycholinguistic properties of words having a slightly higher importance. Owing to its generic nature, the presented approach has the potential to generalize over other exams containing MCQs.
Vocabulary is one of the most important parts of language competence. Testing of vocabulary knowledge is central to research on reading and language. However, it usually costs a large amount of time and human labor to build an item bank and to test large number of students. In this paper, we propose a novel testing strategy by combining automatic item generation (AIG) and computerized adaptive testing (CAT) in vocabulary assessment for Chinese L2 learners. Firstly, we generate three types of vocabulary questions by modeling both the vocabulary knowledge and learners’ writing error data. After evaluation and calibration, we construct a balanced item pool with automatically generated items, and implement a three-parameter computerized adaptive test. We conduct manual item evaluation and online student tests in the experiments. The results show that the combination of AIG and CAT can construct test items efficiently and reduce test cost significantly. Also, the test result of CAT can provide valuable feedback to AIG algorithms.
Computational linguistic research on the language complexity of student writing typically involves human ratings as a gold standard. However, educational science shows that teachers find it difficult to identify and cleanly separate accuracy, different aspects of complexity, contents, and structure. In this paper, we therefore explore the use of computational linguistic methods to investigate how task-appropriate complexity and accuracy relate to the grading of overall performance, content performance, and language performance as assigned by teachers. Based on texts written by students for the official school-leaving state examination (Abitur), we show that teachers successfully assign higher language performance grades to essays with higher task-appropriate language complexity and properly separate this from content scores. Yet, accuracy impacts teacher assessment for all grading rubrics, also the content score, overemphasizing the role of accuracy. Our analysis is based on broad computational linguistic modeling of German language complexity and an innovative theory- and data-driven feature aggregation method inferring task-appropriate language complexity.
We present a model for automatic scoring of coherence based on comparing the rhetorical structure (RS) of college student summaries in L2 (English) against expert summaries. Coherence is conceptualised as a construct consisting of the rhetorical relation and its arguments. Comparison with expert-assigned scores shows that RS scores correlate with both cohesion and coherence. Furthermore, RS scores improve the accuracy of a regression model for cohesion score prediction.
This paper reports on the BEA-2019 Shared Task on Grammatical Error Correction (GEC). As with the CoNLL-2014 shared task, participants are required to correct all types of errors in test data. One of the main contributions of the BEA-2019 shared task is the introduction of a new dataset, the Write&Improve+LOCNESS corpus, which represents a wider range of native and learner English levels and abilities. Another contribution is the introduction of tracks, which control the amount of annotated data available to participants. Systems are evaluated in terms of ERRANT F_0.5, which allows us to report a much wider range of performance statistics. The competition was hosted on Codalab and remains open for further submissions on the blind test set.
Spelling correction has attracted a lot of attention in the NLP community. However, models have been usually evaluated on artificiallycreated or proprietary corpora. A publiclyavailable corpus of authentic misspellings, annotated in context, is still lacking. To address this, we present and release an annotated data set of 6,121 spelling errors in context, based on a corpus of essays written by English language learners. We also develop a minimallysupervised context-aware approach to spelling correction. It achieves strong results on our data: 88.12% accuracy. This approach can also train with a minimal amount of annotated data (performance reduced by less than 1%). Furthermore, this approach allows easy portability to new domains. We evaluate our model on data from a medical domain and demonstrate that it rivals the performance of a model trained and tuned on in-domain data.
The quantity and quality of training data plays a crucial role in grammatical error correction (GEC). However, due to the fact that obtaining human-annotated GEC data is both time-consuming and expensive, several studies have focused on generating artificial error sentences to boost training data for grammatical error correction, and shown significantly better performance. The present study explores how fluency filtering can affect the quality of artificial errors. By comparing artificial data filtered by different levels of fluency, we find that artificial error sentences with low fluency can greatly facilitate error correction, while high fluency errors introduce more noise.
In this paper we present first results for the task of Automated Essay Scoring for Norwegian learner language. We analyze a number of properties of this task experimentally and assess (i) the formulation of the task as either regression or classification, (ii) the use of various non-neural and neural machine learning architectures with various types of input representations, and (iii) applying multi-task learning for joint prediction of essay scoring and native language identification. We find that a GRU-based attention model trained in a single-task setting performs best at the AES task.
Grammatical error detection (GED) in non-native writing requires systems to identify a wide range of errors in text written by language learners. Error detection as a purely supervised task can be challenging, as GED datasets are limited in size and the label distributions are highly imbalanced. Contextualized word representations offer a possible solution, as they can efficiently capture compositional information in language and can be optimized on large amounts of unsupervised data. In this paper, we perform a systematic comparison of ELMo, BERT and Flair embeddings (Peters et al., 2017; Devlin et al., 2018; Akbik et al., 2018) on a range of public GED datasets, and propose an approach to effectively integrate such representations in current methods, achieving a new state of the art on GED. We further analyze the strengths and weaknesses of different contextual embeddings for the task at hand, and present detailed analyses of their impact on different types of errors.
Character-based representations in neural models have been claimed to be a tool to overcome spelling variation in in word token-based input. We examine this claim in neural models for content scoring. We formulate precise hypotheses about the possible effects of adding character representations to word-based models and test these hypotheses on large-scale real world content scoring datasets. We find that, while character representations may provide small performance gains in general, their effectiveness in accounting for spelling variation may be limited. We show that spelling correction can provide larger gains than character representations, and that spelling correction improves the performance of models with character representations. With these insights, we report a new state of the art on the ASAP-SAS content scoring dataset.
Recent work on Grammatical Error Correction (GEC) has highlighted the importance of language modeling in that it is certainly possible to achieve good performance by comparing the probabilities of the proposed edits. At the same time, advancements in language modeling have managed to generate linguistic output, which is almost indistinguishable from that of human-generated text. In this paper, we up the ante by exploring the potential of more sophisticated language models in GEC and offer some key insights on their strengths and weaknesses. We show that, in line with recent results in other NLP tasks, Transformer architectures achieve consistently high performance and provide a competitive baseline for future machine learning models.
We introduce unsupervised techniques based on phrase-based statistical machine translation for grammatical error correction (GEC) trained on a pseudo learner corpus created by Google Translation. We verified our GEC system through experiments on a low resource track of the shared task at BEA2019. As a result, we achieved an F0.5 score of 28.31 points with the test data.
The field of Grammatical Error Correction (GEC) has produced various systems to deal with focused phenomena or general text editing. We propose an automatic way to combine black-box systems. Our method automatically detects the strength of a system or the combination of several systems per error type, improving precision and recall while optimizing F-score directly. We show consistent improvement over the best standalone system in all the configurations tested. This approach also outperforms average ensembling of different RNN models with random initializations. In addition, we analyze the use of BERT for GEC - reporting promising results on this end. We also present a spellchecker created for this task which outperforms standard spellcheckers tested on the task of spellchecking. This paper describes a system submission to Building Educational Applications 2019 Shared Task: Grammatical Error Correction. Combining the output of top BEA 2019 shared task systems using our approach, currently holds the highest reported score in the open phase of the BEA 2019 shared task, improving F-0.5 score by 3.7 points over the best result reported.
It has been demonstrated that the utilization of a monolingual corpus in neural Grammatical Error Correction (GEC) systems can significantly improve the system performance. The previous state-of-the-art neural GEC system is an ensemble of four Transformer models pretrained on a large amount of Wikipedia Edits. The Singsound GEC system follows a similar approach but is equipped with a sophisticated erroneous data generating component. Our system achieved an F0:5 of 66.61 in the BEA 2019 Shared Task: Grammatical Error Correction. With our novel erroneous data generating component, the Singsound neural GEC system yielded an M2 of 63.2 on the CoNLL-2014 benchmark (8.4% relative improvement over the previous state-of-the-art system).
In this paper, we describe two systems we developed for the three tracks we have participated in the BEA-2019 GEC Shared Task. We investigate competitive classification models with bi-directional recurrent neural networks (Bi-RNN) and neural machine translation (NMT) models. For different tracks, we use ensemble systems to selectively combine the NMT models, the classification models, and some rules, and demonstrate that an ensemble solution can effectively improve GEC performance over single systems. Our GEC systems ranked the first in the Unrestricted Track, and the third in both the Restricted Track and the Low Resource Track.
We describe two entries from the Cambridge University Engineering Department to the BEA 2019 Shared Task on grammatical error correction. Our submission to the low-resource track is based on prior work on using finite state transducers together with strong neural language models. Our system for the restricted track is a purely neural system consisting of neural language models and neural machine translation models trained with back-translation and a combination of checkpoint averaging and fine-tuning – without the help of any additional tools like spell checkers. The latter system has been used inside a separate system combination entry in cooperation with the Cambridge University Computer Lab.
We introduce the AIP-Tohoku grammatical error correction (GEC) system for the BEA-2019 shared task in Track 1 (Restricted Track) and Track 2 (Unrestricted Track) using the same system architecture. Our system comprises two key components: error generation and sentence-level error detection. In particular, GEC with sentence-level grammatical error detection is a novel and versatile approach, and we experimentally demonstrate that it significantly improves the precision of the base model. Our system is ranked 9th in Track 1 and 2nd in Track 2.
Our submitted models are NMT systems based on the Transformer model, which we improve by incorporating several enhancements: applying dropout to whole source and target words, weighting target subwords, averaging model checkpoints, and using the trained model iteratively for correcting the intermediate translations. The system in the Restricted Track is trained on the provided corpora with oversampled “cleaner” sentences and reaches 59.39 F0.5 score on the test set. The system in the Low-Resource Track is trained from Wikipedia revision histories and reaches 44.13 F0.5 score. Finally, we finetune the system from the Low-Resource Track on restricted data and achieve 64.55 F0.5 score.
This paper describes our contribution to the low-resource track of the BEA 2019 shared task on Grammatical Error Correction (GEC). Our approach to GEC builds on the theory of the noisy channel by combining a channel model and language model. We generate confusion sets from the Wikipedia edit history and use the frequencies of edits to estimate the channel model. Additionally, we use two pre-trained language models: 1) Google’s BERT model, which we fine-tune for specific error types and 2) OpenAI’s GPT-2 model, utilizing that it can operate with previous sentences as context. Furthermore, we search for the optimal combinations of corrections using beam search.
This paper describes the BLCU Group submissions to the Building Educational Applications (BEA) 2019 Shared Task on Grammatical Error Correction (GEC). The task is to detect and correct grammatical errors that occurred in essays. We participate in 2 tracks including the Restricted Track and the Unrestricted Track. Our system is based on a Transformer model architecture. We integrate many effective methods proposed in recent years. Such as, Byte Pair Encoding, model ensemble, checkpoints average and spell checker. We also corrupt the public monolingual data to further improve the performance of the model. On the test data of the BEA 2019 Shared Task, our system yields F0.5 = 58.62 and 59.50, ranking twelfth and fourth respectively.
We introduce our system that is submitted to the restricted track of the BEA 2019 shared task on grammatical error correction1 (GEC). It is essential to select an appropriate hypothesis sentence from the candidates list generated by the GEC model. A re-ranker can evaluate the naturalness of a corrected sentence using language models trained on large corpora. On the other hand, these language models and language representations do not explicitly take into account the grammatical errors written by learners. Thus, it is not straightforward to utilize language representations trained from a large corpus, such as Bidirectional Encoder Representations from Transformers (BERT), in a form suitable for the learner’s grammatical errors. Therefore, we propose to fine-tune BERT on learner corpora with grammatical errors for re-ranking. The experimental results of the W&I+LOCNESS development dataset demonstrate that re-ranking using BERT can effectively improve the correction performance.
Grammatical error correction can be viewed as a low-resource sequence-to-sequence task, because publicly available parallel corpora are limited.To tackle this challenge, we first generate erroneous versions of large unannotated corpora using a realistic noising function. The resulting parallel corpora are sub-sequently used to pre-train Transformer models. Then, by sequentially applying transfer learning, we adapt these models to the domain and style of the test set. Combined with a context-aware neural spellchecker, our system achieves competitive results in both restricted and low resource tracks in ACL 2019 BEAShared Task. We release all of our code and materials for reproducibility.
In this paper, we describe our submission to the BEA 2019 shared task on grammatical error correction. We present a system pipeline that utilises both error detection and correction models. The input text is first corrected by two complementary neural machine translation systems: one using convolutional networks and multi-task learning, and another using a neural Transformer-based system. Training is performed on publicly available data, along with artificial examples generated through back-translation. The n-best lists of these two machine translation systems are then combined and scored using a finite state transducer (FST). Finally, an unsupervised re-ranking system is applied to the n-best output of the FST. The re-ranker uses a number of error detection features to re-rank the FST n-best list and identify the final 1-best correction hypothesis. Our system achieves 66.75% F 0.5 on error correction (ranking 4th), and 82.52% F 0.5 on token-level error detection (ranking 2nd) in the restricted track of the shared task.
In this paper, we explore two approaches of generating error-focused phrases and examine whether these phrases can lead to better performance in grammatical error correction for the restricted track of BEA 2019 Shared Task on GEC. Our results show that phrases directly extracted from GEC corpora outperform phrases from statistical machine translation phrase table by a large margin. Appending error+context phrases to the original GEC corpora yields comparably high precision. We also explore the generation of artificial syntactic error sentences using error+context phrases for the unrestricted track. The additional training data greatly facilitates syntactic error correction (e.g., verb form) and contributes to better overall performance.
In this paper, we describe our approach to GEC using the BERT model for creation of encoded representation and some of our enhancements, namely, “Heads” are fully-connected networks which are used for finding the errors and later receive recommendation from the networks on dealing with a highlighted part of the sentence only. Among the main advantages of our solution is increasing the system productivity and lowering the time of processing while keeping the high accuracy of GEC results.
Considerable effort has been made to address the data sparsity problem in neural grammatical error correction. In this work, we propose a simple and surprisingly effective unsupervised synthetic error generation method based on confusion sets extracted from a spellchecker to increase the amount of training data. Synthetic data is used to pre-train a Transformer sequence-to-sequence model, which not only improves over a strong baseline trained on authentic error-annotated data, but also enables the development of a practical GEC system in a scenario where little genuine error-annotated data is available. The developed systems placed first in the BEA19 shared task, achieving 69.47 and 64.24 F0.5 in the restricted and low-resource tracks respectively, both on the W&I+LOCNESS test set. On the popular CoNLL 2014 test set, we report state-of-the-art results of 64.16 M² for the submitted system, and 61.30 M² for the constrained system trained on the NUCLE and Lang-8 data.
A number of methods have been proposed to automatically extract collocations, i.e., conventionalized lexical combinations, from text corpora. However, the attempts to evaluate and compare them with a specific application in mind lag behind. This paper compares three end-to-end resources for collocation learning, all of which used the same corpus but different methods. Adopting a gold-standard evaluation method, the results show that the method of dependency parsing outperforms regex-over-pos in collocation identification. The lexical association measures (AMs) used for collocation ranking perform about the same overall but differently for individual collocation types. Further analysis has also revealed that there are considerable differences between other commonly used AMs.
In this paper, we present experiments that estimate the impact of specific lexical choices of people writing in a second language (L2). In particular, we look at misspelled words that indicate lexical uncertainty on the part of the author, and separate them into three categories: misspelled cognates, “L2-ed” (in our case, anglicized) words, and all other spelling errors. We test the assumption that such errors contain clues about the native language of an essay’s author through the task of native language identification. The results of the experiments show that the information brought by each of these categories is complementary. We also note that while the distribution of such features changes with the proficiency level of the writer, their contribution towards native language identification remains significant at all levels.
We present a new concept prerequisite learning method for Learning Object (LO) ordering that exploits only linguistic features extracted from textual educational resources. The method was tested in a cross- and in- domain scenario both for Italian and English. Additionally, we performed experiments based on a incremental training strategy to study the impact of the training set size on the classifier performances. The paper also introduces ITA-PREREQ, to the best of our knowledge the first Italian dataset annotated with prerequisite relations between pairs of educational concepts, and describe the automatic strategy devised to build it.
Existing example retrieval systems do not include grammatically incorrect examples or present only a few examples, if any. Even if a retrieval system has a wide coverage of incorrect examples along with the correct counterpart, learners need to know whether their query includes errors or not. Considering the usability of retrieving incorrect examples, our proposed method uses a large-scale corpus and presents correct expressions along with incorrect expressions using a grammatical error detection system so that the learner do not need to be aware of how to search for the examples. Intrinsic and extrinsic evaluations indicate that our method improves accuracy of example sentence retrieval and quality of learner’s writing.
In this study, we developed an automated algorithm to provide feedback about the specific content of non-native English speakers’ spoken responses. The responses were spontaneous speech, elicited using integrated tasks where the language learners listened to and/or read passages and integrated the core content in their spoken responses. Our models detected the absence of key points considered to be important in a spoken response to a particular test question, based on two different models: (a) a model using word-embedding based content features and (b) a state-of-the art short response scoring engine using traditional n-gram based features. Both models achieved a substantially improved performance over the majority baseline, and the combination of the two models achieved a significant further improvement. In particular, the models were robust to automated speech recognition (ASR) errors, and performance based on the ASR word hypotheses was comparable to that based on manual transcriptions. The accuracy and F-score of the best model for the questions included in the train set were 0.80 and 0.68, respectively. Finally, we discussed possible approaches to generating targeted feedback about the content of a language learner’s response, based on automatically detected missing key points.
This paper provides an analytical assessment of student short answer responses with a view to potential benefits in pedagogical contexts. We first propose and formalize two novel analytical assessment tasks: analytic score prediction and justification identification, and then provide the first dataset created for analytic short answer scoring research. Subsequently, we present a neural baseline model and report our extensive empirical results to demonstrate how our dataset can be used to explore new and intriguing technical challenges in short answer scoring. The dataset is publicly available for research purposes.
Visual content has been proven to be effective for micro-learning compared to other media. In this paper, we discuss leveraging this observation in our efforts to build audio-visual content for young learners’ vocabulary learning. We attempt to tackle two major issues in the process of traditional visual curation tasks. Generic learning videos do not necessarily satisfy the unique context of a learner and/or an educator, and hence may not result in maximal learning outcomes. Also, manual video curation by educators is a highly labor-intensive process. To this end, we present a customizable micro-learning audio-visual content curation tool that is designed to reduce the human (educator) effort in creating just-in-time learning videos from a textual description (learning script). This provides educators with control of the content while preparing the learning scripts, and in turn can also be customized to capture the desired learning objectives and outcomes. As a use case, we automatically generate learning videos with British National Corpus’ (BNC) frequently spoken vocabulary words and evaluate them with experts. They positively recommended the generated learning videos with an average rating of 4.25 on a Likert scale of 5 points. The inter-annotator agreement between the experts for the video quality was substantial (Fleiss Kappa=0.62) with an overall agreement of 81%.
During learning, students often have questions which they would benefit from responses to in real time. In class, a student can ask a question to a teacher. During homework, or even in class if the student is shy, it can be more difficult to receive a rapid response. In this work, we introduce Curio SmartChat, an automated question answering system for middle school Science topics. Our system has now been used by around 20,000 students who have so far asked over 100,000 questions. We present data on the challenge created by students’ grammatical errors and spelling mistakes, and discuss our system’s approach and degree of effectiveness at disambiguating questions that the system is initially unsure about. We also discuss the prevalence of student “small talk” not related to science topics, the pluses and minuses of this behavior, and how a system should respond to these conversational acts. We conclude with discussions and point to directions for potential future work.
This paper discusses the computer-assisted content evaluation of summaries. We propose a method to make a correspondence between the segments of the source text and its summary. As a unit of the segment, we adopt “Idea Unit (IU)” which is proposed in Applied Linguistics. Introducing IUs enables us to make a correspondence even for the sentences that contain multiple ideas. The IU correspondence is made based on the similarity between vector representations of IU. An evaluation experiment with two source texts and 20 summaries showed that the proposed method is more robust against rephrased expressions than the conventional ROUGE-based baselines. Also, the proposed method outperformed the baselines in recall. We im-plemented the proposed method in a GUI tool“Segment Matcher” that aids teachers to estab-lish a link between corresponding IUs acrossthe summary and source text.
Automatic readability assessment aims to ensure that readers read texts that they can comprehend. However, computational models are typically trained on texts created from the perspective of the text writer, not the target reader. There is little experimental research on the relationship between expert annotations of readability, reader’s language proficiency, and different levels of reading comprehension. To address this gap, we conducted a user study in which over a 100 participants read texts of different reading levels and answered questions created to test three forms of comprehension. Our results indicate that more than readability annotation or reader proficiency, it is the type of comprehension question asked that shows differences between reader responses - inferential questions were difficult for users of all levels of proficiency across reading levels. The data collected from this study will be released with this paper, which will, for the first time, provide a collection of 45 reader bench marked texts to evaluate readability assessment systems developed for adult learners of English. It can also potentially be useful for the development of question generation approaches in intelligent tutoring systems research.
The selection of texts for second language learning purposes typically relies on teachers’ and test developers’ individual judgment of the observable qualitative properties of a text. Little or no consideration is generally given to the quantitative dimension within an evidence-based framework of reproducibility. This study aims to fill the gap by evaluating the effectiveness of an automatic tool trained to assess text complexity in the context of Italian as a second language learning. A dataset of texts labeled by expert test developers was used to evaluate the performance of three classifier models (decision tree, random forest, and support vector machine), which were trained using linguistic features measured quantitatively and extracted from the texts. The experimental analysis provided satisfactory results, also in relation to which kind of linguistic trait contributed the most to the final outcome.
We present a machine foreign-language teacher that takes documents written in a student’s native language and detects situations where it can replace words with their foreign glosses such that new foreign vocabulary can be learned simply through reading the resulting mixed-language text. We show that it is possible to design such a machine teacher without any supervised data from (human) students. We accomplish this by modifying a cloze language model to incrementally learn new vocabulary items, and use this language model as a proxy for the word guessing and learning ability of real students. Our machine foreign-language teacher decides which subset of words to replace by consulting this language model. We evaluate three variants of our student proxy language models through a study on Amazon Mechanical Turk (MTurk). We find that MTurk “students” were able to guess the meanings of foreign words introduced by the machine teacher with high accuracy for both function words as well as content words in two out of the three models. In addition, we show that students are able to retain their knowledge about the foreign words after they finish reading the document.
We track the development of writing complexity and accuracy in German students’ early academic language development from first to eighth grade. Combining an empirically broad approach to linguistic complexity with the high-quality error annotation included in the Karlsruhe Children’s Text corpus (Lavalley et al. 2015) used, we construct models of German academic language development that successfully identify the student’s grade level. We show that classifiers for the early years rely more on accuracy development, whereas development in secondary school is better characterized by increasingly complex language in all domains: linguistic system, language use, and human sentence processing characteristics. We demonstrate the generalizability and robustness of models using such a broad complexity feature set across writing topics.
We developed an automated oral proficiency scoring system for non-native English speakers’ spontaneous speech. Automated systems that score holistic proficiency are expected to assess a wide range of performance categories, and the content is one of the core performance categories. In order to assess the quality of the content, we trained a Siamese convolutional neural network (CNN) to model the semantic relationship between key points generated by experts and a test response. The correlation between human scores and Siamese CNN scores was comparable to human-human agreement (r=0.63), and it was higher than the baseline content features. The inclusion of Siamese CNN-based feature to the existing state-of-the-art automated scoring model achieved a small but statistically significant improvement. However, the new model suffered from score inflation for long atypical responses with serious content issues. We investigated the reasons of this score inflation by analyzing the associations with linguistic features and identifying areas strongly associated with the score errors.
A typical medical curriculum is organized in a hierarchy of instructional objectives called Learning Outcomes (LOs); a few thousand LOs span five years of study. Gaining a thorough understanding of the curriculum requires learners to recognize and apply related LOs across years, and across different parts of the curriculum. However, given the large scope of the curriculum, manually labeling related LOs is tedious, and almost impossible to scale. In this paper, we build a system that learns relationships between LOs, and we achieve up to human-level performance in the LO relationship extraction task. We then present an application where the proposed system is employed to build a map of related LOs and Learning Resources (LRs) pertaining to a virtual patient case. We believe that our system can help medical students grasp the curriculum better, within classroom as well as in Intelligent Tutoring Systems (ITS) settings.
This article studies the relationship between text readability indice and automatic machine understanding systems. Our hypothesis is that the simpler a text is, the better it should be understood by a machine. We thus expect to a strong correlation between readability levels on the one hand, and performance of automatic reading systems on the other hand. We test this hypothesis with several understanding systems based on language models of varying strengths, measuring this correlation on two corpora of journalistic texts. Our results suggest that this correlation is rather small that existing comprehension systems are far to reproduce the gradual improvement of their performance on texts of decreasing complexity.
We present an analysis of metaphors in news text simplification. Using features that capture general and metaphor specific characteristics, we test whether we can automatically identify which metaphors will be changed or preserved, and whether there are features that have different predictive power for metaphors or literal words. The experiments show that the Age of Acquisition is the most distinctive feature for both metaphors and literal words. Features that capture Imageability and Concreteness are useful when used alone, but within the full set of features they lose their impact. Frequency of use seems to be the best feature to differentiate metaphors that should be changed and those to be preserved.
This study aims to build an automatic system for the detection of plagiarized spoken responses in the context of an assessment of English speaking proficiency for non-native speakers. Classification models were trained to distinguish between plagiarized and non-plagiarized responses with two different types of features: text-to-text content similarity measures, which are commonly used in the task of plagiarism detection for written documents, and speaking proficiency measures, which were specifically designed for spontaneous speech and extracted using an automated speech scoring system. The experiments were first conducted on a large data set drawn from an operational English proficiency assessment across multiple years, and the best classifier on this heavily imbalanced data set resulted in an F1-score of 0.761 on the plagiarized class. This system was then validated on operational responses collected from a single administration of the assessment and achieved a recall of 0.897. The results indicate that the proposed system can potentially be used to improve the validity of both human and automated assessment of non-native spoken English.
There is a long record of research on equity in schools. As machine learning researchers begin to study fairness and bias in earnest, language technologies in education have an unusually strong theoretical and applied foundation to build on. Here, we introduce concepts from culturally relevant pedagogy and other frameworks for teaching and learning, identifying future work on equity in NLP. We present case studies in a range of topics like intelligent tutoring systems, computer-assisted language learning, automated essay scoring, and sentiment analysis in classrooms, and provide an actionable agenda for research.
Knowing how to use words appropriately has been a key to improving language proficiency. Previous studies typically discuss how students learn receptively to select the correct candidate from a set of confusing words in the fill-in-the-blank task where specific context is given. In this paper, we go one step further, assisting students to learn to use confusing words appropriately in a productive task: sentence translation. We leverage the GiveMe-Example system, which suggests example sentences for each confusing word, to achieve this goal. In this study, students learn to differentiate the confusing words by reading the example sentences, and then choose the appropriate word(s) to complete the sentence translation task. Results show students made substantial progress in terms of sentence structure. In addition, highly proficient students better managed to learn confusing words. In view of the influence of the first language on learners, we further propose an effective approach to improve the quality of the suggested sentences.
One of the challenges of building natural language processing (NLP) applications for education is finding a large domain-specific corpus for the subject of interest (e.g., history or science). To address this challenge, we propose a tool, Dexter, that extracts a subject-specific corpus from a heterogeneous corpus, such as Wikipedia, by relying on a small seed corpus and distributed document representations. We empirically show the impact of the generated corpus on language modeling, estimating word embeddings, and consequently, distractor generation, resulting in better performances than while using a general domain corpus, a heuristically constructed domain-specific corpus, and a corpus generated by a popular system: BootCaT.
In this paper, we investigate the impact of using 4 recent neural models for generating artificial errors to help train the neural grammatical error correction models. We conduct a battery of experiments on the effect of data size, models, and comparison with a rule-based approach.
Automated essay scoring systems typically rely on hand-crafted features to predict essay quality, but such systems are limited by the cost of feature engineering. Neural networks offer an alternative to feature engineering, but they typically require more annotated data. This paper explores network structures, contextualized embeddings and pre-training strategies aimed at capturing discourse characteristics of essays. Experiments on three essay scoring tasks show benefits from all three strategies in different combinations, with simpler architectures being more effective when less training data is available.
Automatic assessment of the proficiency levels of the learner is a critical part of Intelligent Tutoring Systems. We present methods for assessment in the context of language learning. We use a specialized Elo formula used in conjunction with educational data mining. We simultaneously obtain ratings for the proficiency of the learners and for the difficulty of the linguistic concepts that the learners are trying to master. From the same data we also learn a graph structure representing a domain model capturing the relations among the concepts. This application of Elo provides ratings for learners and concepts which correlate well with subjective proficiency levels of the learners and difficulty levels of the concepts.
We present a unique dataset of student source-based argument essays to facilitate research on the relations between content, argumentation skills, and assessment. Two classroom writing assignments were given to college students in a STEM major, accompanied by a carefully designed rubric. The paper presents a reliability study of the rubric, showing it to be highly reliable, and initial annotation on content and argumentation annotation of the essays.
One of the biomedical entity types of relevance for medicine or biosciences are chemical compounds and drugs. The correct detection these entities is critical for other text mining applications building on them, such as adverse drug-reaction detection, medication-related fake news or drug-target extraction. Although a significant effort was made to detect mentions of drugs/chemicals in English texts, so far only very limited attempts were made to recognize them in medical documents in other languages. Taking into account the growing amount of medical publications and clinical records written in Spanish, we have organized the first shared task on detecting drug and chemical entities in Spanish medical documents. Additionally, we included a clinical concept-indexing sub-track asking teams to return SNOMED-CT identifiers related to drugs/chemicals for a collection of documents. For this task, named PharmaCoNER, we generated annotation guidelines together with a corpus of 1,000 manually annotated clinical case studies. A total of 22 teams participated in the sub-track 1, (77 system runs), and 7 teams in the sub-track 2 (19 system runs). Top scoring teams used sophisticated deep learning approaches yielding very competitive results with F-measures above 0.91. These results indicate that there is a real interest in promoting biomedical text mining efforts beyond English. We foresee that the PharmaCoNER annotation guidelines, corpus and participant systems will foster the development of new resources for clinical and biomedical text mining systems of Spanish medical data.
The recognition of pharmacological substances, compounds and proteins is an essential preliminary work for the recognition of relations between chemicals and other biomedically relevant units. In this paper, we describe an approach to Task 1 of the PharmaCoNER Challenge, which involves the recognition of mentions of chemicals and drugs in Spanish medical texts. We train a state-of-the-art BiLSTM-CRF sequence tagger with stacked Pooled Contextualized Embeddings, word and sub-word embeddings using the open-source framework FLAIR. We present a new corpus composed of articles and papers from Spanish health science journals, termed the Spanish Health Corpus, and use it to train domain-specific embeddings which we incorporate in our model training. We achieve a result of 89.76% F1-score using pre-trained embeddings and are able to improve these results to 90.52% F1-score using specialized embeddings.
This paper presents the participation of the VSP team for the PharmaCoNER Tracks from the BioNLP Open Shared Task 2019. The system consists of a neural model for the Named Entity Recognition of drugs, medications and chemical entities in Spanish and the use of the Spanish Edition of SNOMED CT term search engine for the concept normalization of the recognized mentions. The neural network is implemented with two bidirectional Recurrent Neural Networks with LSTM cells that creates a feature vector for each word of the sentences in order to classify the entities. The first layer uses the characters of each word and the resulting vector is aggregated to the second layer together with its word embedding in order to create the feature vector of the word. Besides, a Conditional Random Field layer classifies the vector representation of each word in one of the mention types. The system obtains a performance of 76.29%, and 60.34% in F1 for the classification of the Named Entity Recognition task and the Concept indexing task, respectively. This method presents good results with a basic approach without using pretrained word embeddings or any hand-crafted features.
The aim of this paper is to present our approach (IxaMed) in the PharmacoNER 2019 task. The task consists of identifying chemical, drug, and gene/protein mentions from clinical case studies written in Spanish. The evaluation of the task is divided in two scenarios: one corresponding to the detection of named entities and one corresponding to the indexation of named entities that have been previously identified. In order to identify named entities we have made use of a Bi-LSTM with a CRF on top in combination with different types of word embeddings. We have achieved our best result (86.81 F-Score) combining pretrained word embeddings of Wikipedia and Electronic Health Records (50M words) with contextual string embeddings of Wikipedia and Electronic Health Records. On the other hand, for the indexation of the named entities we have used the Levenshtein distance obtaining a 85.34 F-Score as our best result.
Named entity recognition has been extensively studied on English news texts. However, the transfer to other domains and languages is still a challenging problem. In this paper, we describe the system with which we participated in the first subtrack of the PharmaCoNER competition of the BioNLP Open Shared Tasks 2019. Aiming at pharmacological entity detection in Spanish texts, the task provides a non-standard domain and language setting. However, we propose an architecture that requires neither language nor domain expertise. We treat the task as a sequence labeling task and experiment with attention-based embedding selection and the training on automatically annotated data to further improve our system’s performance. Our system achieves promising results, especially by combining the different techniques, and reaches up to 88.6% F1 in the competition.
The Biological Text Mining Unit at BSC and CNIO organized the first shared task on chemical & drug mention recognition from Spanish medical texts called PharmaCoNER (Pharmacological Substances, Compounds and proteins and Named Entity Recognition track) in 2019, which includes two tracks: one for NER offset and entity classification (track 1) and the other one for concept indexing (track 2). We developed a pipeline system based on deep learning methods for this shared task, specifically, a subsystem based on BERT (Bidirectional Encoder Representations from Transformers) for NER offset and entity classification and a subsystem based on Bpool (Bi-LSTM with max/mean pooling) for concept indexing. Evaluation conducted on the shared task data showed that our system achieves a micro-average F1-score of 0.9105 on track 1 and a micro-average F1-score of 0.8391 on track 2.
In this work, we introduce a Deep Learning architecture for pharmaceutical and chemical Named Entity Recognition in Spanish clinical cases texts. We propose a hybrid model approach based on two Bidirectional Long Short-Term Memory (Bi-LSTM) network and Conditional Random Field (CRF) network using character, word, concept and sense embeddings to deal with the extraction of semantic, syntactic and morphological features. The approach was evaluated on the PharmaCoNER Corpus obtaining an F-measure of 85.24% for subtask 1 and 49.36% for subtask2. These results prove that deep learning methods with specific domain embedding representations can outperform the state-of-the-art approaches.
We present a neural pipeline approach that performs named entity recognition (NER) and concept indexing (CI), which links them to concept unique identifiers (CUIs) in a knowledge base, for the PharmaCoNER shared task on pharmaceutical drugs and chemical entities. We proposed a neural NER model that captures the surrounding semantic information of a given sequence by capturing the forward- and backward-context of bidirectional LSTM (Bi-LSTM) output of a target span using contextual span representation-based exhaustive approach. The NER model enumerates all possible spans as potential entity mentions and classify them into entity types or no entity with deep neural networks. For representing span, we compare several different neural network architectures and their ensembling for the NER model. We then perform dictionary matching for CI and, if there is no matching, we further compute similarity scores between a mention and CUIs using entity embeddings to assign the CUI with the highest score to the mention. We evaluate our approach on the two sub-tasks in the shared task. Among the five submitted runs, the best run for each sub-task achieved the F-score of 86.76% on Sub-task 1 (NER) and the F-score of 79.97% (strict) on Sub-task 2 (CI).
We present the approach of the Turku NLP group to the PharmaCoNER task on Spanish biomedical named entity recognition. We apply a CRF-based baseline approach and multilingual BERT to the task, achieving an F-score of 88% on the development data and 87% on the test set with BERT. Our approach reflects a straightforward application of a state-of-the-art multilingual model that is not specifically tailored to either the language nor the application domain. The source code is available at: https://github.com/chaanim/pharmaconer
The active gene annotation corpus (AGAC) was developed to support knowledge discovery for drug repurposing. Based on the corpus, the AGAC track of the BioNLP Open Shared Tasks 2019 was organized, to facilitate cross-disciplinary collaboration across BioNLP and Pharmacoinformatics communities, for drug repurposing. The AGAC track consists of three subtasks: 1) named entity recognition, 2) thematic relation extraction, and 3) loss of function (LOF) / gain of function (GOF) topic classification. The AGAC track was participated by five teams, of which the performance are compared and analyzed. The the results revealed a substantial room for improvement in the design of the task, which we analyzed in terms of “imbalanced data”, “selective annotation” and “latent topic annotation”.
The prediction of the relationship between the disease with genes and its mutations is a very important knowledge extraction task that can potentially help drug discovery. In this paper, we present our approaches for trigger word detection (task 1) and the identification of its thematic role (task 2) in AGAC track of BioNLP Open Shared Task 2019. Task 1 can be regarded as the traditional name entity recognition (NER), which cultivates molecular phenomena related to gene mutation. Task 2 can be regarded as relation extraction which captures the thematic roles between entities. For two tasks, we exploit the pre-trained biomedical language representation model (i.e., BERT) in the pipe of information extraction for the collection of mutation-disease knowledge from PubMed. And also, we design a fine-tuning technique and extra features by using multi-task learning. The experiment results show that our proposed approaches achieve 0.60 (ranks 1) and 0.25 (ranks 2) on task 1 and task 2 respectively in terms of F1 metric.
Understanding the pathogenesis of genetic diseases through different gene activities and their relations to relevant diseases is important for new drug discovery and drug repositioning. In this paper, we present a joint deep learning model in a multi-task learning paradigm for gene mutation-disease knowledge extraction, DeepGeneMD, which adapts the state-of-the-art hierarchical multi-task learning framework for joint inference on named entity recognition (NER) and relation extraction (RE) in the context of the AGAC (Active Gene Annotation Corpus) track at 2019 BioNLP Open Shared Tasks (BioNLP-OST). It simultaneously extracts gene mutation related activities, diseases, and their relations from the published scientific literature. In DeepGeneMD, we explore the task decomposition to create auxiliary subtasks so that more interactions between different learning subtasks can be leveraged in model training. Our model achieves the average F1 score of 0.45 on recognizing gene activities and disease entities, ranking 2nd in the AGAC NER task; and the average F1 score of 0.35 on extracting relations, ranking 1st in the AGAC RE task.
This paper presents our participation in the AGAC Track from the 2019 BioNLP Open Shared Tasks. We provide a solution for Task 3, which aims to extract “gene - function change - disease” triples, where “gene” and “disease” are mentions of particular genes and diseases respectively and “function change” is one of four pre-defined relationship types. Our system extends BERT (Devlin et al., 2018), a state-of-the-art language model, which learns contextual language representations from a large unlabelled corpus and whose parameters can be fine-tuned to solve specific tasks with minimal additional architecture. We encode the pair of mentions and their textual context as two consecutive sequences in BERT, separated by a special symbol. We then use a single linear layer to classify their relationship into five classes (four pre-defined, as well as ‘no relation’). Despite considerable class imbalance, our system significantly outperforms a random baseline while relying on an extremely simple setup with no specially engineered features.
This paper describes the Named Entity Recognition system of the Institute for Artificial Intelligence “Mihai Drăgănescu” of the Romanian Academy (RACAI for short). Our best F1 score of 0.84984 was achieved using an ensemble of two systems: a gazetteer-based baseline and a RNN-based NER system, developed specially for PharmaCoNER 2019. We will describe the individual systems and the ensemble algorithm, compare the final system to the current state of the art, as well as discuss our results with respect to the quality of the training data and its annotation strategy. The resulting NER system is language independent, provided that language-dependent resources and preprocessing tools exist, such as tokenizers and POS taggers.
To date, a large amount of biomedical content has been published in non-English texts, especially for clinical documents. Therefore, it is of considerable significance to conduct Natural Language Processing (NLP) research in non-English literature. PharmaCoNER is the first Named Entity Recognition (NER) task to recognize chemical and protein entities from Spanish biomedical texts. Since there have been abundant resources in the NLP field, how to exploit these existing resources to a new task to obtain competitive performance is a meaningful study. Inspired by the success of transfer learning with language models, we introduce the BERT benchmark to facilitate the research of PharmaCoNER task. In this paper, we evaluate two baselines based on Multilingual BERT and BioBERT on the PharmaCoNER corpus. Experimental results show that transferring the knowledge learned from source large-scale datasets to the target domain offers an effective solution for the PharmaCoNER task.
This paper presents a novel transfer multi-task learning method for Bacteria Biotope rel+ner task at BioNLP-OST 2019. To alleviate the data deficiency problem in domain-specific information extraction, we use BERT(Bidirectional Encoder Representations from Transformers) and pre-train it using mask language models and next sentence prediction on both general corpus and medical corpus like PubMed. In fine-tuning stage, we fine-tune the relation extraction layer and mention recognition layer designed by us on the top of BERT to extract mentions and relations simultaneously. The evaluation results show that our method achieves the best performance on all metrics (including slot error rate, precision and recall) in the Bacteria Biotope rel+ner subtask.
We participated in the BioNLP 2019 Open Shared Tasks: binary relation extraction of SeeDev task. The model was constructed us- ing convolutional neural networks (CNN) and long short term memory networks (LSTM). The full text information and context information were collected using the advantages of CNN and LSTM. The model consisted of two main modules: distributed semantic representation construction, such as word embedding, distance embedding and entity type embed- ding; and CNN-LSTM model. The F1 value of our participated task on the test data set of all types was 0.342. We achieved the second highest in the task. The results showed that our proposed method performed effectively in the binary relation extraction.
In this paper we describe a new named entity extraction system. Our work proposes a system for the identification and annotation of drug names in Spanish biomedical texts based on machine learning and deep learning models. Subsequently, a standardized code using Snomed is assigned to these drugs, for this purpose, Natural Language Processing tools and techniques have been used, and a dictionary of different sources of information has been built. The results are promising, we obtain 78% in F1 score on the first sub-track and in the second task we map with Snomed correctly 72% of the found entities.
This paper presents the fourth edition of the Bacteria Biotope task at BioNLP Open Shared Tasks 2019. The task focuses on the extraction of the locations and phenotypes of microorganisms from PubMed abstracts and full-text excerpts, and the characterization of these entities with respect to reference knowledge sources (NCBI taxonomy, OntoBiotope ontology). The task is motivated by the importance of the knowledge on biodiversity for fundamental research and applications in microbiology. The paper describes the different proposed subtasks, the corpus characteristics, and the challenge organization. We also provide an analysis of the results obtained by participants, and inspect the evolution of the results since the last edition in 2016.
Named Entity Recognition (NER) and Relation Extraction (RE) are essential tools in distilling knowledge from biomedical literature. This paper presents our findings from participating in BioNLP Shared Tasks 2019. We addressed Named Entity Recognition including nested entities extraction, Entity Normalization and Relation Extraction. Our proposed approach of Named Entities can be generalized to different languages and we have shown it’s effectiveness for English and Spanish text. We investigated linguistic features, hybrid loss including ranking and Conditional Random Fields (CRF), multi-task objective and token level ensembling strategy to improve NER. We employed dictionary based fuzzy and semantic search to perform Entity Normalization. Finally, our RE system employed Support Vector Machine (SVM) with linguistic features. Our NER submission (team:MIC-CIS) ranked first in BB-2019 norm+NER task with standard error rate (SER) of 0.7159 and showed competitive performance on PharmaCo NER task with F1-score of 0.8662. Our RE system ranked first in the SeeDev-binary Relation Extraction Task with F1-score of 0.3738.
Different representations of the same concept could often be seen in scientific reports and publications. Entity normalization (or entity linking) is the task to match the different representations to their standard concepts. In this paper, we present a two-step ensemble CNN method that normalizes microbiology-related entities in free text to concepts in standard dictionaries. The method is capable of linking entities when only a small microbiology-related biomedical corpus is available for training, and achieved reasonable performance in the online test of the BioNLP-OST19 shared task Bacteria Biotope.
This paper presents our participation to the Bacteria Biotope Task of the BioNLP Shared Task 2019. Our participation includes two systems for the two subtasks of the Bacteria Biotope Task: the normalization of entities (BB-norm) and the identification of the relations between the entities given a biomedical text (BB-rel). For the normalization of entities, we utilized word embeddings and syntactic re-ranking. For the relation extraction task, pre-defined rules are used. Although both approaches are unsupervised, in the sense that they do not need any labeled data, they achieved promising results. Especially, for the BB-norm task, the results have shown that the proposed method performs as good as deep learning based methods, which require labeled data.
abstract In this article, we describe our approach for the Bacteria Biotopes relation extraction (BB-rel) subtask in the BioNLP Shared Task 2019. This task aims to promote the development of text mining systems that extract relationships between Microorganism, Habitat and Phenotype entities. In this paper, we propose a novel approach for dependency graph construction based on lexical chains, so one dependency graph can represent one or multiple sentences. After that, we propose a neural network model which consists of the bidirectional long short-term memories and an attention graph convolution neural network to learn relation extraction features from the graph. Our approach is able to extract both intra- and inter-sentence relations, and meanwhile utilize syntax information. The results show that our approach achieved the best F1 (66.3%) in the official evaluation participated by 7 teams.
In this paper, we present our participation in the Bacteria Biotope (BB) task at BioNLP-OST 2019. Our system utilizes fine-tuned language representation models and machine learning approaches based on word embedding and lexical features for entities recognition, normalization and relation extraction. It achieves the state-of-the-art performance and is among the top two systems in five of all six subtasks.
As part of the BioNLP Open Shared Tasks 2019, the CRAFT Shared Tasks 2019 provides a platform to gauge the state of the art for three fundamental language processing tasks — dependency parse construction, coreference resolution, and ontology concept identification — over full-text biomedical articles. The structural annotation task requires the automatic generation of dependency parses for each sentence of an article given only the article text. The coreference resolution task focuses on linking coreferring base noun phrase mentions into chains using the symmetrical and transitive identity relation. The ontology concept annotation task involves the identification of concept mentions within text using the classes of ten distinct ontologies in the biomedical domain, both unmodified and augmented with extension classes. This paper provides an overview of each task, including descriptions of the data provided to participants and the evaluation metrics used, and discusses participant results relative to baseline performances for each of the three tasks.
As our submission to the CRAFT shared task 2019, we present two neural approaches to concept recognition. We propose two different systems for joint named entity recognition (NER) and normalization (NEN), both of which model the task as a sequence labeling problem. Our first system is a BiLSTM network with two separate outputs for NER and NEN trained from scratch, whereas the second system is an instance of BioBERT fine-tuned on the concept-recognition task. We exploit two strategies for extending concept coverage, ontology pretraining and backoff with a dictionary lookup. Our results show that the backoff strategy effectively tackles the problem of unseen concepts, addressing a major limitation of the chosen design. In the cross-system comparison, BioBERT proves to be a strong basis for creating a concept-recognition system, although some entity types are predicted more accurately by the BiLSTM-based system.
This paper describes our system developed for the coreference resolution task of the CRAFT Shared Tasks 2019. The CRAFT corpus is more challenging than other existing corpora because it contains full text articles. We have employed an existing span-based state-of-theart neural coreference resolution system as a baseline system. We enhance the system with two different techniques to capture longdistance coreferent pairs. Firstly, we filter noisy mentions based on parse trees with increasing the number of antecedent candidates. Secondly, instead of relying on the LSTMs, we integrate the highly expressive language model–BERT into our model. Experimental results show that our proposed systems significantly outperform the baseline. The best performing system obtained F-scores of 44%, 48%, 39%, 49%, 40%, and 57% on the test set with B3, BLANC, CEAFE, CEAFM, LEA, and MUC metrics, respectively. Additionally, the proposed model is able to detect coreferent pairs in long distances, even with a distance of more than 200 sentences.
We present the approach taken by the TurkuNLP group in the CRAFT Structural Annotation task, a shared task on dependency parsing. Our approach builds primarily on the Turku neural parser, a native dependency parser that ranked among the best in the recent CoNLL tasks on parsing Universal Dependencies. To adapt the parser to the biomedical domain, we considered and evaluated a number of approaches, including the generation of custom word embeddings, combination with other in-domain resources, and the incorporation of information from named entity recognition. We achieved a labeled attachment score of 89.7%, the best result among task participants.
BioNLP Open Shared Tasks (BioNLP-OST) is an international competition organized to facilitate development and sharing of computational tasks of biomedical text mining and solutions to them. For BioNLP-OST 2019, we introduced a new mental health informatics task called “RDoC Task”, which is composed of two subtasks: information retrieval and sentence extraction through National Institutes of Mental Health’s Research Domain Criteria framework. Five and four teams around the world participated in the two tasks, respectively. According to the performance on the two tasks, we observe that there is room for improvement for text mining on brain research and mental illness.
This paper presents our system details and results of participation in the RDoC Tasks of BioNLP-OST 2019. Research Domain Criteria (RDoC) construct is a multi-dimensional and broad framework to describe mental health disorders by combining knowledge from genomics to behaviour. Non-availability of RDoC labelled dataset and tedious labelling process hinders the use of RDoC framework to reach its full potential in Biomedical research community and Healthcare industry. Therefore, Task-1 aims at retrieval and ranking of PubMed abstracts relevant to a given RDoC construct and Task-2 aims at extraction of the most relevant sentence from a given PubMed abstract. We investigate (1) attention based supervised neural topic model and SVM for retrieval and ranking of PubMed abstracts and, further utilize BM25 and other relevance measures for re-ranking, (2) supervised and unsupervised sentence ranking models utilizing multi-view representations comprising of query-aware attention-based sentence representation (QAR), bag-of-words (BoW) and TF-IDF. Our best systems achieved 1st rank and scored 0.86 mAP and 0.58 macro average accuracy in Task-1 and Task-2 respectively.
Assessing how individuals perform different activities is key information for modeling health states of individuals and populations. Descriptions of activity performance in clinical free text are complex, including syntactic negation and similarities to textual entailment tasks. We explore a variety of methods for the novel task of classifying four types of assertions about activity performance: Able, Unable, Unclear, and None (no information). We find that ensembling an SVM trained with lexical features and a CNN achieves 77.9% macro F1 score on our task, and yields nearly 80% recall on the rare Unclear and Unable samples. Finally, we highlight several challenges in classifying performance assertions, including capturing information about sources of assistance, incorporating syntactic structure and negation scope, and handling new modalities at test time. Our findings establish a strong baseline for this novel task, and identify intriguing areas for further research.
The objective of this work is to develop an automated diagnosis system that is able to predict the probability of appendicitis given a free-text emergency department (ED) note and additional structured information (e.g., lab test results). Our clinical corpus consists of about 180,000 ED notes based on ten years of patient visits to the Accident and Emergency (A&E) Department of the National University Hospital (NUH), Singapore. We propose a novel neural network approach that learns to diagnose acute appendicitis based on doctors’ free-text ED notes without any feature engineering. On a test set of 2,000 ED notes with equal number of appendicitis (positive) and non-appendicitis (negative) diagnosis and in which all the negative ED notes only consist of abdominal-related diagnosis, our model is able to achieve a promising F_0.5-score of 0.895 while ED doctors achieve F_0.5-score of 0.900. Visualization shows that our model is able to learn important features, signs, and symptoms of patients from unstructured free-text ED notes, which will help doctors to make better diagnosis.
This paper proposes a dataset and method for automatically generating paraphrases for clinical questions relating to patient-specific information in electronic health records (EHRs). Crowdsourcing is used to collect 10,578 unique questions across 946 semantically distinct paraphrase clusters. This corpus is then used with a deep learning-based question paraphrasing method utilizing variational autoencoder and LSTM encoder/decoder. The ultimate use of such a method is to improve the performance of automatic question answering methods for EHRs.
Systematic comparison of methods for relation extraction (RE) is difficult because many experiments in the field are not described precisely enough to be completely reproducible and many papers fail to report ablation studies that would highlight the relative contributions of their various combined techniques. In this work, we build a unifying framework for RE, applying this on three highly used datasets (from the general, biomedical and clinical domains) with the ability to be extendable to new datasets. By performing a systematic exploration of modeling, pre-processing and training methodologies, we find that choices of preprocessing are a large contributor performance and that omission of such information can further hinder fair comparison. Other insights from our exploration allow us to provide recommendations for future research in this area.
Despite recent advances in the application of deep neural networks to various kinds of medical data, extracting information from unstructured textual sources remains a challenging task. The challenges of training and interpreting document classification models are amplified when dealing with small and highly technical datasets, as are common in the clinical domain. Using a dataset of de-identified clinical letters gathered at a memory clinic, we construct several recurrent neural network models for letter classification, and evaluate them on their ability to build meaningful representations of the documents and predict patients’ diagnoses. Additionally, we probe sentence embedding models in order to build a human-interpretable representation of the neural network’s features, using a simple and intuitive technique based on perturbative approaches to sentence importance. In addition to showing which sentences in a document are most informative about the patient’s condition, this method reveals the types of sentences that lead the model to make incorrect diagnoses. Furthermore, we identify clusters of sentences in the embedding space that correlate strongly with importance scores for each clinical diagnosis class.
Inspired by the success of the General Language Understanding Evaluation benchmark, we introduce the Biomedical Language Understanding Evaluation (BLUE) benchmark to facilitate research in the development of pre-training language representations in the biomedicine domain. The benchmark consists of five tasks with ten datasets that cover both biomedical and clinical texts with different dataset sizes and difficulties. We also evaluate several baselines based on BERT and ELMo and find that the BERT model pre-trained on PubMed abstracts and MIMIC-III clinical notes achieves the best results. We make the datasets, pre-trained models, and codes publicly available at https://github.com/ ncbi-nlp/BLUE_Benchmark.
The goal of this work is to utilize Electronic Medical Record (EMR) data for real-time Clinical Decision Support (CDS). We present a deep learning approach to combining in real time available diagnosis codes (ICD codes) and free-text notes: Patient Context Vectors. Patient Context Vectors are created by averaging ICD code embeddings, and by predicting the same from free-text notes via a Convolutional Neural Network. The Patient Context Vectors were then simply appended to available structured data (vital signs and lab results) to build prediction models for a specific condition. Experiments on predicting ARDS, a rare and complex condition, demonstrate the utility of Patient Context Vectors as a means of summarizing the patient history and overall condition, and improve significantly the prediction model results.
In an era when large amounts of data are generated daily in various fields, the biomedical field among others, linguistic resources can be exploited for various tasks of Natural Language Processing. Moreover, increasing number of biomedical documents are available in languages other than English. To be able to extract information from natural language free text resources, methods and tools are needed for a variety of languages. This paper presents the creation of the MoNERo corpus, a gold standard biomedical corpus for Romanian, annotated with both part of speech tags and named entities. MoNERo comprises 154,825 morphologically annotated tokens and 23,188 entity annotations belonging to four entity semantic groups corresponding to UMLS Semantic Groups.
Domain adaptation remains one of the most challenging aspects in the wide-spread use of Semantic Role Labeling (SRL) systems. Current state-of-the-art methods are typically trained on large-scale datasets, but their performances do not directly transfer to low-resource domain-specific settings. In this paper, we propose two approaches for domain adaptation in the biological domain that involves pre-training LSTM-CRF based on existing large-scale datasets and adapting it for a low-resource corpus of biological processes. Our first approach defines a mapping between the source labels and the target labels, and the other approach modifies the final CRF layer in sequence-labeling neural network architecture. We perform our experiments on ProcessBank dataset which contains less than 200 paragraphs on biological processes. We improve over the previous state-of-the-art system on this dataset by 21 F1 points. We also show that, by incorporating event-event relationship in ProcessBank, we are able to achieve an additional 2.6 F1 gain, giving us possible insights into how to improve SRL systems for biological process using richer annotations.
Automatic identification and expansion of ambiguous abbreviations are essential for biomedical natural language processing applications, such as information retrieval and question answering systems. In this paper, we present DEep Contextualized Biomedical Abbreviation Expansion (DECBAE) model. DECBAE automatically collects substantial and relatively clean annotated contexts for 950 ambiguous abbreviations from PubMed abstracts using a simple heuristic. Then it utilizes BioELMo to extract the contextualized features of words, and feed those features to abbreviation-specific bidirectional LSTMs, where the hidden states of the ambiguous abbreviations are used to assign the exact definitions. Our DECBAE model outperforms other baselines by large margins, achieving average accuracy of 0.961 and macro-F1 of 0.917 on the dataset. It also surpasses human performance for expanding a sample abbreviation, and remains robust in imbalanced, low-resources and clinical settings.
Patients and their families often require a better understanding of medical information provided by doctors. We currently address this issue by improving the identification of difficult to understand medical words. We introduce novel embeddings received from RNN - FrnnMUTE (French RNN Medical Understandability Text Embeddings) which allow to reach up to 87.0 F1 score in identification of difficult words. We also note that adding pre-trained FastText word embeddings to the feature set substantially improves the performance of the model which classifies words according to their difficulty. We study the generalizability of different models through three cross-validation scenarios which allow testing classifiers in real-world conditions: understanding of medical words by new users, and classification of new unseen words by the automatic models. The RNN - FrnnMUTE embeddings and the categorization code are being made available for the research.
Systematic reviews are important in evidence based medicine, but are expensive to produce. Automating or semi-automating the data extraction of index test, target condition, and reference standard from articles has the potential to decrease the cost of conducting systematic reviews of diagnostic test accuracy, but relevant training data is not available. We create a distantly supervised dataset of approximately 90,000 sentences, and let two experts manually annotate a small subset of around 1,000 sentences for evaluation. We evaluate the performance of BioBERT and logistic regression for ranking the sentences, and compare the performance for distant and direct supervision. Our results suggest that distant supervision can work as well as, or better than direct supervision on this problem, and that distantly trained models can perform as well as, or better than human annotators.
In this paper, we address the problem of automatically constructing a relevant corpus of scientific articles about food-drug interactions. There is a growing number of scientific publications that describe food-drug interactions but currently building a high-coverage corpus that can be used for information extraction purposes is not trivial. We investigate several methods for automating the query selection process using an expert-curated corpus of food-drug interactions. Our experiments show that index term features along with a decision tree classifier are the best approach for this task and that feature selection approaches and in particular gain ratio outperform frequency-based methods for query selection.
Verbs play a fundamental role in many biomed-ical tasks and applications such as relation and event extraction. We hypothesize that performance on many downstream tasks can be improved by aligning the input pretrained embeddings according to semantic verb classes.In this work, we show that by using semantic clusters for verbs, a large lexicon of verbclasses derived from biomedical literature, weare able to improve the performance of common pretrained embeddings in downstream tasks by retrofitting them to verb classes. We present a simple and computationally efficient approach using a widely-available “off-the-shelf” retrofitting algorithm to align pretrained embeddings according to semantic verb clusters. We achieve state-of-the-art results on text classification and relation extraction tasks.
Distributed representations of text can be used as features when training a statistical classifier. These representations may be created as a composition of word vectors or as context-based sentence vectors. We compare the two kinds of representations (word versus context) for three classification problems: influenza infection classification, drug usage classification and personal health mention classification. For statistical classifiers trained for each of these problems, context-based representations based on ELMo, Universal Sentence Encoder, Neural-Net Language Model and FLAIR are better than Word2Vec, GloVe and the two adapted using the MESH ontology. There is an improvement of 2-4% in the accuracy when these context-based representations are used instead of word-based representations.
Knowledge base construction is crucial for summarising, understanding and inferring relationships between biomedical entities. However, for many practical applications such as drug discovery, the scarcity of relevant facts (e.g. gene X is therapeutic target for disease Y) severely limits a domain expert’s ability to create a usable knowledge base, either directly or by training a relation extraction model. In this paper, we present a simple and effective method of extracting new facts with a pre-specified binary relationship type from the biomedical literature, without requiring any training data or hand-crafted rules. Our system discovers, ranks and presents the most salient patterns to domain experts in an interpretable form. By marking patterns as compatible with the desired relationship type, experts indirectly batch-annotate candidate pairs whose relationship is expressed with such patterns in the literature. Even with a complete absence of seed data, experts are able to discover thousands of high-quality pairs with the desired relationship within minutes. When a small number of relevant pairs do exist - even when their relationship is more general (e.g. gene X is biologically associated with disease Y) than the relationship of interest - our system leverages them in order to i) learn a better ranking of the patterns to be annotated or ii) generate weakly labelled pairs in a fully automated manner. We evaluate our method both intrinsically and via a downstream knowledge base completion task, and show that it is an effective way of constructing knowledge bases when few or no relevant facts are already available.
We report the work-in-progress of collecting MedLexSp, an unified medical lexicon for the Spanish language, featuring terms and inflected word forms mapped to Unified Medical Language System (UMLS) Concept Unique Identifiers (CUIs), semantic types and groups. First, we leveraged a list of term lemmas and forms from a previous project, and mapped them to UMLS terms and CUIs. To enrich the lexicon, we used both domain-corpora (e.g. Summaries of Product Characteristics and MedlinePlus) and natural language processing techniques such as string distance methods or generation of syntactic variants of multi-word terms. We also added term variants by mapping their CUIs to missing items available in the Spanish versions of standard thesauri (e.g. Medical Subject Headings and World Health Organization Adverse Drug Reactions terminology). We enhanced the vocabulary coverage by gathering missing terms from resources such as the Anatomical Therapeutical Classification, the National Cancer Institute (NCI) Dictionary of Cancer Terms, OrphaData, or the Nomenclátor de Prescripción for drug names. Part-of-Speech information is being included in the lexicon, and the current version amounts up to 76 454 lemmas and 203 043 inflected forms (including conjugated verbs, number and gender variants), corresponding to 30 647 UMLS CUIs. MedLexSp is distributed freely for research purposes.
The goal of text classification is to automatically assign categories to documents. Deep learning automatically learns effective features from data instead of adopting human-designed features. In this paper, we focus specifically on biomedical document classification using a deep learning approach. We present a novel multichannel TextCNN model for MeSH term indexing. Beyond the normal use of the text from the abstract and title for model training, we also consider figure and table captions, as well as paragraphs associated with the figures and tables. We demonstrate that these latter text sources are important feature sources for our method. A new dataset consisting of these text segments curated from 257,590 full text articles together with the articles’ MEDLINE/PubMed MeSH terms is publicly available.
Automatic extraction of relations and interactions between biological entities from scientific literature remains an extremely challenging problem in biomedical information extraction and natural language processing in general. One of the reasons for slow progress is the relative scarcity of standardized and publicly available benchmarks. In this paper we introduce BioRelEx, a new dataset of fully annotated sentences from biomedical literature that capture binding interactions between proteins and/or biomolecules. To foster reproducible research on the interaction extraction task, we define a precise and transparent evaluation process, tools for error analysis and significance tests. Finally, we conduct extensive experiments to evaluate several baselines, including SciIE, a recently introduced neural multi-task architecture that has demonstrated state-of-the-art performance on several tasks.
This paper describes a natural language processing (NLP) approach to extracting lactation-specific drug information from two sources: FDA-mandated drug labels and the NLM Drugs and Lactation Database (LactMed). A frame semantic approach is utilized, and the paper describes the selected frames, their annotation on a set of 900 sections from drug labels and LactMed articles, and the NLP system to extract such frame instances automatically. The ultimate goal of the project is to use such a system to identify discrepancies in lactation-related drug information between these resources.
To automatically analyse complex trajectory information enclosed in clinical text (e.g. timing of symptoms, duration of treatment), it is important to understand the related temporal aspects, anchoring each event on an absolute point in time. In the clinical domain, few temporally annotated corpora are currently available. Moreover, underlying annotation schemas - which mainly rely on the TimeML standard - are not necessarily easily applicable for applications such as patient timeline reconstruction. In this work, we investigated how temporal information is documented in clinical text by annotating a corpus of medical reports with time expressions (TIMEXes), based on TimeML. The developed corpus is available to the NLP community. Starting from our annotations, we analysed the suitability of the TimeML TIMEX schema for capturing timeline information, identifying challenges and possible solutions. As a result, we propose a novel annotation schema that could be useful for timeline reconstruction: CALendar EXpression (CALEX).
The automatic processing of clinical documents, such as Electronic Health Records (EHRs), could benefit substantially from the enrichment of medical terminologies with terms encountered in clinical practice. To integrate such terms into existing knowledge sources, they must be linked to corresponding concepts. We present a method for the semantic categorization of clinical terms based on their surface form. We find that features based on sublanguage properties can provide valuable cues for the classification of term variants.
In this paper, we presented an improved methodology to extract PIO elements, from abstracts of medical papers, that reduces ambiguity. The proposed technique was used to build a dataset of PIO elements that we call PICONET. We further proposed a model of PIO elements classification using state of the art BERT embedding. In addition, we investigated a contextualized embedding, BioBERT, trained on medical corpora. It has been found that using the BioBERT embedding improved the classification accuracy, outperforming the BERT-based model. This result reinforces the idea of the importance of embedding contextualization in subsequent classification tasks in this specific context.Furthermore, to enhance the accuracy of the model, we have investigated an ensemble method based on the LGBM algorithm. We trained the LGBM model, with the above models as base learners, to learn a linear combination of the predicted probabilities for the 3 classes with the TF-IDF score and the QIEF that optimizes the classification. The results indicate that these text features were good features to consider in order to boost the deeply contextualized classification model. We compared the performance of the classifier when using the features with one of the base learners and the case where we combine the base learners along with the features. We obtained the highest score in terms of AUC when we combine the base learners.The present work resulted in the creation of a PIO element dataset, PICONET, and a classification tool. These constitute and important component of our system of automatic mining of medical abstracts. We intend to extend the dataset to full medical articles. The model will be modified to take into account the higher complexity of full text data and more efficient features for model boosting will be investigated.
Having in mind that different languages might present different challenges, this paper presents the following contributions to the area of Information Extraction from clinical text, targeting the Portuguese language: a collection of 281 clinical texts in this language, with manually-annotated named entities; word embeddings trained in a larger collection of similar texts; results of using BiLSTM-CRF neural networks for named entity recognition on the annotated collection, including a comparison of using in-domain or out-of-domain word embeddings in this task. Although learned with much less data, performance is higher when using in-domain embeddings. When tested in 20 independent clinical texts, this model achieved better results than a model using larger out-of-domain embeddings.
We present two models for combining word and character embeddings for cause-of-death classification of verbal autopsy reports using the text of the narratives. We find that for smaller datasets (500 to 1000 records), adding character information to the model improves classification, making character-based CNNs a promising method for automated verbal autopsy coding.
A major obstacle to the development of Natural Language Processing (NLP) methods in the biomedical domain is data accessibility. This problem can be addressed by generating medical data artificially. Most previous studies have focused on the generation of short clinical text, and evaluation of the data utility has been limited. We propose a generic methodology to guide the generation of clinical text with key phrases. We use the artificial data as additional training data in two key biomedical NLP tasks: text classification and temporal relation extraction. We show that artificially generated training data used in conjunction with real training data can lead to performance boosts for data-greedy neural network algorithms. We also demonstrate the usefulness of the generated data for NLP setups where it fully replaces real training data.
Question answering (QA) is a challenging task in natural language processing (NLP), especially when it is applied to specific domains. While models trained in the general domain can be adapted to a new target domain, their performance often degrades significantly due to domain mismatch. Alternatively, one can require a large amount of domain-specific QA data, but such data are rare, especially for the medical domain. In this study, we first collect a large-scale Chinese medical QA corpus called ChiMed; second we annotate a small fraction of the corpus to check the quality of the answers; third, we extract two datasets from the corpus and use them for the relevancy prediction task and the adoption prediction task. Several benchmark models are applied to the datasets, producing good results for both tasks.
The text of clinical notes can be a valuable source of patient information and clinical assessments. Historically, the primary approach for exploiting clinical notes has been information extraction: linking spans of text to concepts in a detailed domain ontology. However, recent work has demonstrated the potential of supervised machine learning to extract document-level codes directly from the raw text of clinical notes. We propose to bridge the gap between the two approaches with two novel syntheses: (1) treating extracted concepts as features, which are used to supplement or replace the text of the note; (2) treating extracted concepts as labels, which are used to learn a better representation of the text. Unfortunately, the resulting concepts do not yield performance gains on the document-level clinical coding task. We explore possible explanations and future research directions.
Textual data are useful for accessing expert information. Yet, since the texts are representative of distinct language uses, it is necessary to build specific corpora in order to be able to design suitable NLP tools. In some domains, such as medical domain, it may be complicated to access the representative textual data and their semantic annotations, while there exists a real need for providing efficient tools and methods. Our paper presents a corpus of clinical cases written in French, and their semantic annotations. Thus, we manually annotated a set of 717 files into four general categories (age, gender, outcome, and origin) for a total number of 2,835 annotations. The values of age, gender, and outcome are normalized. A subset with 70 files has been additionally manually annotated into 27 categories for a total number of 5,198 annotations.
A large percentage of medical information is in unstructured text format in electronic medical record systems. Manual extraction of information from clinical notes is extremely time consuming. Natural language processing has been widely used in recent years for automatic information extraction from medical texts. However, algorithms trained on data from a single healthcare provider are not generalizable and error-prone due to the heterogeneity and uniqueness of medical documents. We develop a two-stage federated natural language processing method that enables utilization of clinical notes from different hospitals or clinics without moving the data, and demonstrate its performance using obesity and comorbities phenotyping as medical task. This approach not only improves the quality of a specific clinical task but also facilitates knowledge progression in the whole healthcare system, which is an essential part of learning health system. To the best of our knowledge, this is the first application of federated machine learning in clinical NLP.
We consider the task of detecting sentences that express causality, as a step towards mining causal relations from texts. To bypass the scarcity of causal instances in relation extraction datasets, we exploit transfer learning, namely ELMO and BERT, using a bidirectional GRU with self-attention ( BIGRUATT ) as a baseline. We experiment with both generic public relation extraction datasets and a new biomedical causal sentence detection dataset, a subset of which we make publicly available. We find that transfer learning helps only in very small datasets. With larger datasets, BIGRUATT reaches a performance plateau, then larger datasets and transfer learning do not help.
Network Embedding (NE) methods, which map network nodes to low-dimensional feature vectors, have wide applications in network analysis and bioinformatics. Many existing NE methods rely only on network structure, overlooking other information associated with the nodes, e.g., text describing the nodes. Recent attempts to combine the two sources of information only consider local network structure. We extend NODE2VEC, a well-known NE method that considers broader network structure, to also consider textual node descriptors using recurrent neural encoders. Our method is evaluated on link prediction in two networks derived from UMLS. Experimental results demonstrate the effectiveness of the proposed approach compared to previous work.
The purpose of automatic text simplification is to transform technical or difficult to understand texts into a more friendly version. The semantics must be preserved during this transformation. Automatic text simplification can be done at different levels (lexical, syntactic, semantic, stylistic...) and relies on the corresponding knowledge and resources (lexicon, rules...). Our objective is to propose methods and material for the creation of transformation rules from a small set of parallel sentences differentiated by their technicity. We also propose a typology of transformations and quantify them. We work with French-language data related to the medical domain, although we assume that the method can be exploited on texts in any language and from any domain.
Despite recent advances in natural language processing, many statistical models for processing text perform extremely poorly under domain shift. Processing biomedical and clinical text is a critically important application area of natural language processing, for which there are few robust, practical, publicly available models. This paper describes scispaCy, a new Python library and models for practical biomedical/scientific text processing, which heavily leverages the spaCy library. We detail the performance of two packages of models released in scispaCy and demonstrate their robustness on several tasks and datasets. Models and code are available at https://allenai.github.io/scispacy/.
Chemical patents are an important resource for chemical information. However, few chemical Named Entity Recognition (NER) systems have been evaluated on patent documents, due in part to their structural and linguistic complexity. In this paper, we explore the NER performance of a BiLSTM-CRF model utilising pre-trained word embeddings, character-level word representations and contextualized ELMo word representations for chemical patents. We compare word embeddings pre-trained on biomedical and chemical patent corpora. The effect of tokenizers optimized for the chemical domain on NER performance in chemical patents is also explored. The results on two patent corpora show that contextualized word representations generated from ELMo substantially improve chemical NER performance w.r.t. the current state-of-the-art. We also show that domain-specific resources such as word embeddings trained on chemical patents and chemical-specific tokenizers, have a positive impact on NER performance.
The availability of large-scale and real-time data on social media has motivated research into adverse drug reactions (ADRs). ADR classification helps to identify negative effects of drugs, which can guide health professionals and pharmaceutical companies in making medications safer and advocating patients’ safety. Based on the observation that in social media, negative sentiment is frequently expressed towards ADRs, this study presents a neural model that combines sentiment analysis with transfer learning techniques to improve ADR detection in social media postings. Our system is firstly trained to classify sentiment in tweets concerning current affairs, using the SemEval17-task4A corpus. We then apply transfer learning to adapt the model to the task of detecting ADRs in social media postings. We show that, in combination with rich representations of words and their contexts, transfer learning is beneficial, especially given the large degree of vocabulary overlap between the current affairs posts in the SemEval17-task4A corpus and posts about ADRs. We compare our results with previous approaches, and show that our model can outperform them by up to 3% F-score.
In research best practices can change over time as new discoveries are made and novel methods are implemented. Scientific publications reporting about the latest facts and current state-of-the-art can be possibly outdated after some years or even proved to be false. A publication usually sheds light only on the knowledge of the period it has been published. Thus, the aspect of time can play an essential role in the reliability of the presented information. In Natural Language Processing many methods focus on information extraction from text, such as detecting entities and their relationship to each other. Those methods mostly focus on the facts presented in the text itself and not on the aspects of knowledge which changes over time. This work instead examines the evolution in biomedical knowledge over time using scientific literature in terms of diachronic change. Mainly the usage of temporal and distributional concept representations are explored and evaluated by a proof-of-concept.
Randomized controlled trials assess the effects of an experimental intervention by comparing it to a control intervention with regard to some variables - trial outcomes. Statistical hypothesis testing is used to test if the experimental intervention is superior to the control. Statistical significance is typically reported for the measured outcomes and is an important characteristic of the results. We propose a machine learning approach to automatically extract reported outcomes, significance levels and the relation between them. We annotated a corpus of 663 sentences with 2,552 outcome - significance level relations (1,372 positive and 1,180 negative relations). We compared several classifiers, using a manually crafted feature set, and a number of deep learning models. The best performance (F-measure of 94%) was shown by the BioBERT fine-tuned model.
This paper presents the MEDIQA 2019 shared task organized at the ACL-BioNLP workshop. The shared task is motivated by a need to develop relevant methods, techniques and gold standards for inference and entailment in the medical domain, and their application to improve domain specific information retrieval and question answering systems. MEDIQA 2019 includes three tasks: Natural Language Inference (NLI), Recognizing Question Entailment (RQE), and Question Answering (QA) in the medical domain. 72 teams participated in the challenge, achieving an accuracy of 98% in the NLI task, 74.9% in the RQE task, and 78.3% in the QA task. In this paper, we describe the tasks, the datasets, and the participants’ approaches and results. We hope that this shared task will attract further research efforts in textual inference, question entailment, and question answering in the medical domain.
This paper describes the models designated for the MEDIQA 2019 shared tasks by the team PANLP. We take advantages of the recent advances in pre-trained bidirectional transformer language models such as BERT (Devlin et al., 2018) and MT-DNN (Liu et al., 2019b). We find that pre-trained language models can significantly outperform traditional deep learning models. Transfer learning from the NLI task to the RQE task is also experimented, which proves to be useful in improving the results of fine-tuning MT-DNN large. A knowledge distillation process is implemented, to distill the knowledge contained in a set of models and transfer it into an single model, whose performance turns out to be comparable with that obtained by the ensemble of that set of models. Finally, for test submissions, model ensemble and a re-ranking process are implemented to boost the performances. Our models participated in all three tasks and ranked the 1st place for the RQE task, and the 2nd place for the NLI task, and also the 2nd place for the QA task.
Parallel deep learning architectures like fine-tuned BERT and MT-DNN, have quickly become the state of the art, bypassing previous deep and shallow learning methods by a large margin. More recently, pre-trained models from large related datasets have been able to perform well on many downstream tasks by just fine-tuning on domain-specific datasets (similar to transfer learning). However, using powerful models on non-trivial tasks, such as ranking and large document classification, still remains a challenge due to input size limitations of parallel architecture and extremely small datasets (insufficient for fine-tuning). In this work, we introduce an end-to-end system, trained in a multi-task setting, to filter and re-rank answers in the medical domain. We use task-specific pre-trained models as deep feature extractors. Our model achieves the highest Spearman’s Rho and Mean Reciprocal Rank of 0.338 and 0.9622 respectively, on the ACL-BioNLP workshop MediQA Question Answering shared-task.
This paper describes our competing system to enter the MEDIQA-2019 competition. We use a multi-source transfer learning approach to transfer the knowledge from MT-DNN and SciBERT to natural language understanding tasks in the medical domain. For transfer learning fine-tuning, we use multi-task learning on NLI, RQE and QA tasks on general and medical domains to improve performance. The proposed methods are proved effective for natural language understanding in the medical domain, and we rank the first place on the QA task.
While deep learning techniques have shown promising results in many natural language processing (NLP) tasks, it has not been widely applied to the clinical domain. The lack of large datasets and the pervasive use of domain-specific language (i.e. abbreviations and acronyms) in the clinical domain causes slower progress in NLP tasks than that of the general NLP tasks. To fill this gap, we employ word/subword-level based models that adopt large-scale data-driven methods such as pre-trained language models and transfer learning in analyzing text for the clinical domain. Empirical results demonstrate the superiority of the proposed methods by achieving 90.6% accuracy in medical domain natural language inference task. Furthermore, we inspect the independent strengths of the proposed approaches in quantitative and qualitative manners. This analysis will help researchers to select necessary components in building models for the medical domain.
Natural language inference (NLI) is challenging, especially when it is applied to technical domains such as biomedical settings. In this paper, we propose a hybrid approach to biomedical NLI where different types of information are exploited for this task. Our base model includes a pre-trained text encoder as the core component, and a syntax encoder and a feature encoder to capture syntactic and domain-specific information. Then we combine the output of different base models to form more powerful ensemble models. Finally, we design two conflict resolution strategies when the test data contain multiple (premise, hypothesis) pairs with the same premise. We train our models on the MedNLI dataset, yielding the best performance on the test set of the MEDIQA 2019 Task 1.
In this paper, we describe our system and results submitted for the Natural Language Inference (NLI) track of the MEDIQA 2019 Shared Task. As KU_ai team, we used BERT as our baseline model and pre-processed the MedNLI dataset to mitigate the negative impact of de-identification artifacts. Moreover, we investigated different pre-training and transfer learning approaches to improve the performance. We show that pre-training the language model on rich biomedical corpora has a significant effect in teaching the model domain-specific language. In addition, training the model on large NLI datasets such as MultiNLI and SNLI helps in learning task-specific reasoning. Finally, we ensembled our highest-performing models, and achieved 84.7% accuracy on the unseen test dataset and ranked 10th out of 17 teams in the official results.
In this paper, we propose a novel model called Adversarial Multi-Task Network (AMTN) for jointly modeling Recognizing Question Entailment (RQE) and medical Question Answering (QA) tasks. AMTN utilizes a pre-trained BioBERT model and an Interactive Transformer to learn the shared semantic representations across different task through parameter sharing mechanism. Meanwhile, an adversarial training strategy is introduced to separate the private features of each task from the shared representations. Experiments on BioNLP 2019 RQE and QA Shared Task datasets show that our model benefits from the shared representations of both tasks provided by multi-task learning and adversarial training, and obtains significant improvements upon the single-task models.
In medical domain, given a medical question, it is difficult to manually select the most relevant information from a large number of search results. BioNLP 2019 proposes Question Answering (QA) task, which encourages the use of text mining technology to automatically judge whether a search result is an answer to the medical question. The main challenge of QA task is how to mine the semantic relation between question and answer. We propose BioBERT Transformer model to tackle this challenge, which applies Transformers to extract semantic relation between different words in questions and answers. Furthermore, BioBERT is utilized to encode medical domain-specific contextualized word representations. Our method has reached the accuracy of 76.24% and spearman of 17.12% on the BioNLP 2019 QA task.
This paper presents the submissions by TeamDr.Quad to the ACL-BioNLP 2019 shared task on Textual Inference and Question Entailment in the Medical Domain. Our system is based on the prior work Liu et al. (2019) which uses a multi-task objective function for textual entailment. In this work, we explore different strategies for generalizing state-of-the-art language understanding models to the specialized medical domain. Our results on the shared task demonstrate that incorporating domain knowledge through data augmentation is a powerful strategy for addressing challenges posed specialized domains such as medicine.
This paper presents a multi-task learning approach to natural language inference (NLI) and question entailment (RQE) in the biomedical domain. Recognizing textual inference relations and question similarity can address the issue of answering new consumer health questions by mapping them to Frequently Asked Questions on reputed websites like the NIH. We show that leveraging information from parallel tasks across domains along with medical knowledge integration allows our model to learn better biomedical feature representations. Our final models for the NLI and RQE tasks achieve the 4th and 2nd rank on the shared-task leaderboard respectively.
Official System Description paper of Team IIT-KGP ranked 1st in the Development phase and 3rd in Testing Phase in MEDIQA 2019 - Recognizing Question Entailment (RQE) Shared Task of BioNLP workshop - ACL 2019. The number of people turning to the Internet to search for a diverse range of health-related subjects continues to grow and with this multitude of information available, duplicate questions are becoming more frequent and finding the most appropriate answers becomes problematic. This issue is important for question answering platforms as it complicates the retrieval of all information relevant to the same topic, particularly when questions similar in essence are expressed differently, and answering a given medical question by retrieving similar questions that are already answered by human experts seems to be a promising solution. In this paper, we present our novel approach to detect question entailment by determining the type of question asked rather than focusing on the type of the ailment given. This unique methodology makes the approach robust towards examples which have different ailment names but are synonyms of each other. Also, it enables us to check entailment at a much more fine-grained level. QSpider is a staged system consisting of state-of-the-art model Sci-BERT used as a multi-class classifier aimed at capturing both question types and semantic relations stacked with a Gradient Boosting Classifier which checks for entailment. QSpider achieves an accuracy score of 68.4% on the Test set which outperforms the baseline model (54.1%) by an accuracy score of 14.3%.
We report on our system for textual inference and question entailment in the medical domain for the ACL BioNLP 2019 Shared Task, MEDIQA. Textual inference is the task of finding the semantic relationships between pairs of text. Question entailment involves identifying pairs of questions which have similar semantic content. To improve upon medical natural language inference and question entailment approaches to further medical question answering, we propose a system that incorporates open-domain and biomedical domain approaches to improve semantic understanding and ambiguity resolution. Our models achieve 80% accuracy on medical natural language inference (6.5% absolute improvement over the original baseline), 48.9% accuracy on recognising medical question entailment, 0.248 Spearman’s rho for question answering ranking and 68.6% accuracy for question answering classification.
In this paper, we present Biomedical Multi-Task Deep Neural Network (Bio-MTDNN) on the NLI task of MediQA 2019 challenge. Bio-MTDNN utilizes “transfer learning” based paradigm where not only the source and target domains are different but also the source and target tasks are varied, although related. Further, Bio-MTDNN integrates knowledge from external sources such as clinical databases (UMLS) enhancing its performance on the clinical domain. Our proposed method outperformed the official baseline and other prior models (such as ESIM and Infersent on dev set) by a considerable margin as evident from our experimental results.
This article describes the participation of the UU_TAILS team in the 2019 MEDIQA challenge intended to improve domain-specific models in medical and clinical NLP. The challenge consists of 3 tasks: medical language inference (NLI), recognizing textual entailment (RQE) and question answering (QA). Our team participated in tasks 1 and 2 and our best runs achieved a performance accuracy of 0.852 and 0.584 respectively for the test sets. The models proposed for task 1 relied on BERT embeddings and different ensemble techniques. For the RQE task, we trained a traditional multilayer perceptron network based on embeddings generated by the universal sentence encoder.
Recent advances in distributed language modeling have led to large performance increases on a variety of natural language processing (NLP) tasks. However, it is not well understood how these methods may be augmented by knowledge-based approaches. This paper compares the performance and internal representation of an Enhanced Sequential Inference Model (ESIM) between three experimental conditions based on the representation method: Bidirectional Encoder Representations from Transformers (BERT), Embeddings of Semantic Predications (ESP), or Cui2Vec. The methods were evaluated on the Medical Natural Language Inference (MedNLI) subtask of the MEDIQA 2019 shared task. This task relied heavily on semantic understanding and thus served as a suitable evaluation set for the comparison of these representation methods.
Natural Language inference is the task of identifying relation between two sentences as entailment, contradiction or neutrality. MedNLI is a biomedical flavour of NLI for clinical domain. This paper explores the use of Bidirectional Encoder Representation from Transformer (BERT) for solving MedNLI. The proposed model, BERT pre-trained on PMC, PubMed and fine-tuned on MIMICIII v1.4, achieves state of the art results on MedNLI (83.45%) and an accuracy of 78.5% in MEDIQA challenge. The authors present an analysis of the attention patterns that emerged as a result of training BERT on MedNLI using a visualization tool, bertviz.
This paper presents the experiments accomplished as a part of our participation in the MEDIQA challenge, an (Abacha et al., 2019) shared task. We participated in all the three tasks defined in this particular shared task. The tasks are viz. i. Natural Language Inference (NLI) ii. Recognizing Question Entailment(RQE) and their application in medical Question Answering (QA). We submitted runs using multiple deep learning based systems (runs) for each of these three tasks. We submitted five system results in each of the NLI and RQE tasks, and four system results for the QA task. The systems yield encouraging results in all the three tasks. The highest performance obtained in NLI, RQE and QA tasks are 81.8%, 53.2%, and 71.7%, respectively.
Biomedical Question Answering (QA) aims at providing automated answers to user questions, regarding a variety of biomedical topics. For example, these questions may ask for related to diseases, drugs, symptoms, or medical procedures. Automated biomedical QA systems could improve the retrieval of information necessary to answer these questions. The MEDIQA challenge consisted of three tasks concerning various aspects of biomedical QA. This challenge aimed at advancing approaches to Natural Language Inference (NLI) and Recognizing Question Entailment (RQE), which would then result in enhanced approaches to biomedical QA. Our approach explored a common Transformer-based architecture that could be applied to each task. This approach shared the same pre-trained weights, but which were then fine-tuned for each task using the provided training data. Furthermore, we augmented the training data with external datasets and enriched the question and answer texts using MER, a named entity recognition tool. Our approach obtained high levels of accuracy, in particular on the NLI task, which classified pairs of text according to their relation. For the QA task, we obtained higher Spearman’s rank correlation values using the entities recognized by MER.
This study describes the model design of the NCUEE system for the MEDIQA challenge at the ACL-BioNLP 2019 workshop. We use the BERT (Bidirectional Encoder Representations from Transformers) as the word embedding method to integrate the BiLSTM (Bidirectional Long Short-Term Memory) network with an attention mechanism for medical text inferences. A total of 42 teams participated in natural language inference task at MEDIQA 2019. Our best accuracy score of 0.84 ranked the top-third among all submissions in the leaderboard.
In this paper, we present three approaches for Natural Language Inference, Question Entailment Recognition and Question-Answering to improve domain-specific Information Retrieval. For addressing the NLI task, the UMLS Metathesaurus was used to find the synonyms of medical terms in given sentences, on which the InferSent model was trained to predict if the given sentence is an entailment, contradictory or neutral. We also introduce a new Extreme gradient boosting model built on PubMed embeddings to perform RQE. Further, a closed-domain Question Answering technique that uses Bi-directional LSTMs trained on the SquAD dataset to determine relevant ranks of answers for a given question is also discussed.
While sequence-to-sequence models have shown remarkable generalization power across several natural language tasks, their construct of solutions are argued to be less compositional than human-like generalization. In this paper, we present seq2attn, a new architecture that is specifically designed to exploit attention to find compositional patterns in the input. In seq2attn, the two standard components of an encoder-decoder model are connected via a transcoder, that modulates the information flow between them. We show that seq2attn can successfully generalize, without requiring any additional supervision, on two tasks which are specifically constructed to challenge the compositional skills of neural networks. The solutions found by the model are highly interpretable, allowing easy analysis of both the types of solutions that are found and potential causes for mistakes. We exploit this opportunity to introduce a new paradigm to test compositionality that studies the extent to which a model overgeneralizes when confronted with exceptions. We show that seq2attn exhibits such overgeneralization to a larger degree than a standard sequence-to-sequence model.
Neural methods for sentiment analysis have led to quantitative improvements over previous approaches, but these advances are not always accompanied with a thorough analysis of the qualitative differences. Therefore, it is not clear what outstanding conceptual challenges for sentiment analysis remain. In this work, we attempt to discover what challenges still prove a problem for sentiment classifiers for English and to provide a challenging dataset. We collect the subset of sentences that an (oracle) ensemble of state-of-the-art sentiment classifiers misclassify and then annotate them for 18 linguistic and paralinguistic phenomena, such as negation, sarcasm, modality, etc. Finally, we provide a case study that demonstrates the usefulness of the dataset to probe the performance of a given sentiment classifier with respect to linguistic phenomena.
We simulate first- and second-order context overlap and show that Skip-Gram with Negative Sampling is similar to Singular Value Decomposition in capturing second-order co-occurrence information, while Pointwise Mutual Information is agnostic to it. We support the results with an empirical study finding that the models react differently when provided with additional second-order information. Our findings reveal a basic property of Skip-Gram with Negative Sampling and point towards an explanation of its success on a variety of tasks.
Monotonicity reasoning is one of the important reasoning skills for any intelligent natural language inference (NLI) model in that it requires the ability to capture the interaction between lexical and syntactic structures. Since no test set has been developed for monotonicity reasoning with wide coverage, it is still unclear whether neural models can perform monotonicity reasoning in a proper way. To investigate this issue, we introduce the Monotonicity Entailment Dataset (MED). Performance by state-of-the-art NLI models on the new test set is substantially worse, under 55%, especially on downward reasoning. In addition, analysis using a monotonicity-driven data augmentation method showed that these models might be limited in their generalization ability in upward and downward reasoning.
Self-explaining text categorization requires a classifier to make a prediction along with supporting evidence. A popular type of evidence is sub-sequences extracted from the input text which are sufficient for the classifier to make the prediction. In this work, we define multi-granular ngrams as basic units for explanation, and organize all ngrams into a hierarchical structure, so that shorter ngrams can be reused while computing longer ngrams. We leverage the tree-structured LSTM to learn a context-independent representation for each unit via parameter sharing. Experiments on medical disease classification show that our model is more accurate, efficient and compact than the BiLSTM and CNN baselines. More importantly, our model can extract intuitive multi-granular evidence to support its predictions.
The correct interpretation of quantifier statements in the context of a visual scene requires non-trivial inference mechanisms. For the example of “most”, we discuss two strategies which rely on fundamentally different cognitive concepts. Our aim is to identify what strategy deep learning models for visual question answering learn when trained on such questions. To this end, we carefully design data to replicate experiments from psycholinguistics where the same question was investigated for humans. Focusing on the FiLM visual question answering model, our experiments indicate that a form of approximate number system emerges whose performance declines with more difficult scenes as predicted by Weber’s law. Moreover, we identify confounding factors, like spatial arrangement of the scene, which impede the effectiveness of this system.
Work on “learning with rationales” shows that humans providing explanations to a machine learning system can improve the system’s predictive accuracy. However, this work has not been connected to work in “explainable AI” which concerns machines explaining their reasoning to humans. In this work, we show that learning with rationales can also improve the quality of the machine’s explanations as evaluated by human judges. Specifically, we present experiments showing that, for CNN-based text classification, explanations generated using “supervised attention” are judged superior to explanations generated using normal unsupervised attention.
The Transformer is a fully attention-based alternative to recurrent networks that has achieved state-of-the-art results across a range of NLP tasks. In this paper, we analyze the structure of attention in a Transformer language model, the GPT-2 small pretrained model. We visualize attention for individual instances and analyze the interaction between attention and syntax over a large corpus. We find that attention targets different parts of speech at different layer depths within the model, and that attention aligns with dependency relations most strongly in the middle layers. We also find that the deepest layers of the model capture the most distant relationships. Finally, we extract exemplar sentences that reveal highly specific patterns targeted by particular attention heads.
Language is a powerful tool which can be used to state the facts as well as express our views and perceptions. Most of the times, we find a subtle bias towards or against someone or something. When it comes to politics, media houses and journalists are known to create bias by shrewd means such as misinterpreting reality and distorting viewpoints towards some parties. This misinterpretation on a large scale can lead to the production of biased news and conspiracy theories. Automating bias detection in newspaper articles could be a good challenge for research in NLP. We proposed a headline attention network for this bias detection. Our model has two distinctive characteristics: (i) it has a structure that mirrors a person’s way of reading a news article (ii) it has attention mechanism applied on the article based on its headline, enabling it to attend to more critical content to predict bias. As the required datasets were not available, we created a dataset comprising of 1329 news articles collected from various Telugu newspapers and marked them for bias towards a particular political party. The experiments conducted on it demonstrated that our model outperforms various baseline methods by a substantial margin.
Neural network models have been very successful in natural language inference, with the best models reaching 90% accuracy in some benchmarks. However, the success of these models turns out to be largely benchmark specific. We show that models trained on a natural language inference dataset drawn from one benchmark fail to perform well in others, even if the notion of inference assumed in these benchmarks is the same or similar. We train six high performing neural network models on different datasets and show that each one of these has problems of generalizing when we replace the original test set with a test set taken from another corpus designed for the same task. In light of these results, we argue that most of the current neural network models are not able to generalize well in the task of natural language inference. We find that using large pre-trained language models helps with transfer learning when the datasets are similar enough. Our results also highlight that the current NLI datasets do not cover the different nuances of inference extensively enough.
Character-level models have been used extensively in recent years in NLP tasks as both supplements and replacements for closed-vocabulary token-level word representations. In one popular architecture, character-level LSTMs are used to feed token representations into a sequence tagger predicting token-level annotations such as part-of-speech (POS) tags. In this work, we examine the behavior of POS taggers across languages from the perspective of individual hidden units within the character LSTM. We aggregate the behavior of these units into language-level metrics which quantify the challenges that taggers face on languages with different morphological properties, and identify links between synthesis and affixation preference and emergent behavior of the hidden tagger layer. In a comparative experiment, we show how modifying the balance between forward and backward hidden units affects model arrangement and performance in these types of languages.
AI systems’ ability to explain their reasoning is critical to their utility and trustworthiness. Deep neural networks have enabled significant progress on many challenging problems such as visual question answering (VQA). However, most of them are opaque black boxes with limited explanatory capability. This paper presents a novel approach to developing a high-performing VQA system that can elucidate its answers with integrated textual and visual explanations that faithfully reflect important aspects of its underlying reasoning while capturing the style of comprehensible human explanations. Extensive experimental evaluation demonstrates the advantages of this approach compared to competing methods using both automated metrics and human evaluation.
Recently, several methods have been proposed to explain the predictions of recurrent neural networks (RNNs), in particular of LSTMs. The goal of these methods is to understand the network’s decisions by assigning to each input variable, e.g., a word, a relevance indicating to which extent it contributed to a particular prediction. In previous works, some of these methods were not yet compared to one another, or were evaluated only qualitatively. We close this gap by systematically and quantitatively comparing these methods in different settings, namely (1) a toy arithmetic task which we use as a sanity check, (2) a five-class sentiment prediction of movie reviews, and besides (3) we explore the usefulness of word relevances to build sentence-level representations. Lastly, using the method that performed best in our experiments, we show how specific linguistic phenomena such as the negation in sentiment analysis reflect in terms of relevance patterns, and how the relevance visualization can help to understand the misclassification of individual samples.
We present a detailed comparison of two types of sequence to sequence models trained to conduct a compositional task. The models are architecturally identical at inference time, but differ in the way that they are trained: our baseline model is trained with a task-success signal only, while the other model receives additional supervision on its attention mechanism (Attentive Guidance), which has shown to be an effective method for encouraging more compositional solutions. We first confirm that the models with attentive guidance indeed infer more compositional solutions than the baseline, by training them on the lookup table task presented by Liska et al. (2019). We then do an in-depth analysis of the structural differences between the two model types, focusing in particular on the organisation of the parameter space and the hidden layer activations and find noticeable differences in both these aspects. Guided networks focus more on the components of the input rather than the sequence as a whole and develop small functional groups of neurons with specific purposes that use their gates more selectively. Results from parameter heat maps, component swapping and graph analysis also indicate that guided networks exhibit a more modular structure with a small number of specialized, strongly connected neurons.
The generalized Dyck language has been used to analyze the ability of Recurrent Neural Networks (RNNs) to learn context-free grammars (CFGs). Recent studies draw conflicting conclusions on their performance, especially regarding the generalizability of the models with respect to the depth of recursion. In this paper, we revisit several common models and experimental settings, discuss the potential problems of the tasks and analyses. Furthermore, we explore the use of attention mechanisms within the seq2seq framework to learn the Dyck language, which could compensate for the limited encoding ability of RNNs. Our findings reveal that attention mechanisms still cannot truly generalize over the recursion depth, although they perform much better than other models on the closing bracket tagging task. Moreover, this also suggests that this commonly used task is not sufficient to test a model’s understanding of CFGs.
A common approach in knowledge base completion (KBC) is to learn representations for entities and relations in order to infer missing facts by generalizing existing ones. A shortcoming of standard models is that they do not explain their predictions to make them verifiable easily to human inspection. In this paper, we propose the Context Path Model (CPM) which generates explanations for new facts in KBC by providing sets of context paths as supporting evidence for these triples. For example, a new triple (Theresa May, nationality, Britain) may be explained by the path (Theresa May, born in, Eastbourne, contained in, Britain). The CPM is formulated as a wrapper that can be applied on top of various existing KBC models. We evaluate it for the well-established TransE model. We observe that its performance remains very close despite the added complexity, and that most of the paths proposed as explanations provide meaningful evidence to assess the correctness.
The recent wide-spread and strong interest in RNNs has spurred detailed investigations of the distributed representations they generate and specifically if they exhibit properties similar to those characterising human languages. Results are at present inconclusive. In this paper, we extend previous work on long-distance dependencies in three ways. We manipulate word embeddings to translate them in a space that is attuned to the linguistic properties under study. We extend the work to sentence embeddings and to new languages. We confirm previous negative results: word embeddings and sentence embeddings do not unequivocally encode fine-grained linguistic properties of long-distance dependencies.
Derivation is a type of a word-formation process which creates new words from existing ones by adding, changing or deleting affixes. In this paper, we explore the potential of word embeddings to identify properties of word derivations in the morphologically rich Czech language. We extract derivational relations between pairs of words from DeriNet, a Czech lexical network, which organizes almost one million Czech lemmas into derivational trees. For each such pair, we compute the difference of the embeddings of the two words, and perform unsupervised clustering of the resulting vectors. Our results show that these clusters largely match manually annotated semantic categories of the derivational relations (e.g. the relation ‘bake–baker’ belongs to category ‘actor’, and a correct clustering puts it into the same cluster as ‘govern–governor’).
Work using artificial languages as training input has shown that LSTMs are capable of inducing the stack-like data structures required to represent context-free and certain mildly context-sensitive languages — formal language classes which correspond in theory to the hierarchical structures of natural language. Here we present a suite of experiments probing whether neural language models trained on linguistic data induce these stack-like data structures and deploy them while incrementally predicting words. We study two natural language phenomena: center embedding sentences and syntactic island constraints on the filler–gap dependency. In order to properly predict words in these structures, a model must be able to temporarily suppress certain expectations and then recover those expectations later, essentially pushing and popping these expectations on a stack. Our results provide evidence that models can successfully suppress and recover expectations in many cases, but do not fully recover their previous grammatical state.
In this paper, we define and apply representational stability analysis (ReStA), an intuitive way of analyzing neural language models. ReStA is a variant of the popular representational similarity analysis (RSA) in cognitive neuroscience. While RSA can be used to compare representations in models, model components, and human brains, ReStA compares instances of the same model, while systematically varying single model parameter. Using ReStA, we study four recent and successful neural language models, and evaluate how sensitive their internal representations are to the amount of prior context. Using RSA, we perform a systematic study of how similar the representational spaces in the first and second (or higher) layers of these models are to each other and to patterns of activation in the human brain. Our results reveal surprisingly strong differences between language models, and give insights into where the deep linguistic processing, that integrates information over multiple sentences, is happening in these models. The combination of ReStA and RSA on models and brains allows us to start addressing the important question of what kind of linguistic processes we can hope to observe in fMRI brain imaging data. In particular, our results suggest that the data on story reading from Wehbe et al./ (2014) contains a signal of shallow linguistic processing, but show no evidence on the more interesting deep linguistic processing.
We propose a novel approach to the study of how artificial neural network perceive the distinction between grammatical and ungrammatical sentences, a crucial task in the growing field of synthetic linguistics. The method is based on performance measures of language models trained on corpora and fine-tuned with either grammatical or ungrammatical sentences, then applied to (different types of) grammatical or ungrammatical sentences. The results show that both in the difficult and highly symmetrical task of detecting subject islands and in the more open CoLA dataset, grammatical sentences give rise to better scores than ungrammatical ones, possibly because they can be better integrated within the body of linguistic structural knowledge that the language model has accumulated.
The quality of Neural Machine Translation (NMT) has been shown to significantly degrade when confronted with source-side noise. We present the first large-scale study of state-of-the-art English-to-German NMT on real grammatical noise, by evaluating on several Grammar Correction corpora. We present methods for evaluating NMT robustness without true references, and we use them for extensive analysis of the effects that different grammatical errors have on the NMT output. We also introduce a technique for visualizing the divergence distribution caused by a source-side error, which allows for additional insights.
Neural network architectures have been augmented with differentiable stacks in order to introduce a bias toward learning hierarchy-sensitive regularities. It has, however, proven difficult to assess the degree to which such a bias is effective, as the operation of the differentiable stack is not always interpretable. In this paper, we attempt to detect the presence of latent representations of hierarchical structure through an exploration of the unsupervised learning of constituency structure. Using a technique due to Shen et al. (2018a,b), we extract syntactic trees from the pushing behavior of stack RNNs trained on language modeling and classification objectives. We find that our models produce parses that reflect natural language syntactic constituencies, demonstrating that stack RNNs do indeed infer linguistically relevant hierarchical structure.
In this paper, we propose a white-box attack algorithm called “Global Search” method and compare it with a simple misspelling noise and a more sophisticated and common white-box attack approach called “Greedy Search”. The attack methods are evaluated on the Convolutional Neural Network (CNN) sentiment classifier trained on the IMDB movie review dataset. The attack success rate is used to evaluate the effectiveness of the attack methods and the perplexity of the sentences is used to measure the degree of distortion of the generated adversarial examples. The experiment results show that the proposed “Global Search” method generates more powerful adversarial examples with less distortion or less modification to the source text.
How and to what extent does BERT encode syntactically-sensitive hierarchical information or positionally-sensitive linear information? Recent work has shown that contextual representations like BERT perform well on tasks that require sensitivity to linguistic structure. We present here two studies which aim to provide a better understanding of the nature of BERT’s representations. The first of these focuses on the identification of structurally-defined elements using diagnostic classifiers, while the second explores BERT’s representation of subject-verb agreement and anaphor-antecedent dependencies through a quantitative assessment of self-attention vectors. In both cases, we find that BERT encodes positional information about word tokens well on its lower layers, but switches to a hierarchically-oriented encoding on higher layers. We conclude then that BERT’s representations do indeed model linguistically relevant aspects of hierarchical structure, though they do not appear to show the sharp sensitivity to hierarchical structure that is found in human processing of reflexive anaphora.
This paper presents a simple but general and effective method to debug the output of machine learning (ML) supervised models, including neural networks. The algorithm looks for features that lower the evaluation metric in such a way that it cannot be ascribed to chance (as measured by their p-values). Using this method – implemented as MLEval tool – you can find: (1) anomalies in test sets, (2) issues in preprocessing, (3) problems in the ML model itself. It can give you an insight into what can be improved in the datasets and/or the model. The same method can be used to compare ML models or different versions of the same model. We present the tool, the theory behind it and use cases for text-based models of various types.
We inspect the multi-head self-attention in Transformer NMT encoders for three source languages, looking for patterns that could have a syntactic interpretation. In many of the attention heads, we frequently find sequences of consecutive states attending to the same position, which resemble syntactic phrases. We propose a transparent deterministic method of quantifying the amount of syntactic information present in the self-attentions, based on automatically building and evaluating phrase-structure trees from the phrase-like sequences. We compare the resulting trees to existing constituency treebanks, both manually and by computing precision and recall.
Large pre-trained neural networks such as BERT have had great recent success in NLP, motivating a growing body of research investigating what aspects of language they are able to learn from unlabeled data. Most recent analysis has focused on model outputs (e.g., language model surprisal) or internal vector representations (e.g., probing classifiers). Complementary to these works, we propose methods for analyzing the attention mechanisms of pre-trained models and apply them to BERT. BERT’s attention heads exhibit patterns such as attending to delimiter tokens, specific positional offsets, or broadly attending over the whole sentence, with heads in the same layer often exhibiting similar behaviors. We further show that certain attention heads correspond well to linguistic notions of syntax and coreference. For example, we find heads that attend to the direct objects of verbs, determiners of nouns, objects of prepositions, and coreferent mentions with remarkably high accuracy. Lastly, we propose an attention-based probing classifier and use it to further demonstrate that substantial syntactic information is captured in BERT’s attention.
The paper presents an unsupervised method for quickly extending a Ukrainian lexicon by generating paradigms and morphological feature structures for new Named Entities and neologisms, which are not covered by existing static morphological resources. This approach addresses a practical problem of modelling paradigms for entities created by the dynamic processes in the lexicon: this problem is especially serious for highly-inflected languages in domains with specialised or quickly changing lexicon. The method uses an unannotated Ukrainian corpus and a small fixed set of inflection tables, which can be found in traditional grammar textbooks. The advantage of the proposed approach is that updating the morphological lexicon does not require training or linguistic annotation, allowing fast knowledge-light extension of an existing static lexicon to improve morphological coverage on a specific corpus. The method is implemented in an open-source package on a GitHub repository. It can be applied to other low-resourced inflectional languages which have internet corpora and linguistic descriptions of their inflection system, following the example of inflection tables for Ukrainian. Evaluation results shows consistent improvements in coverage for Ukrainian corpora of different corpus types.
We present our work on the problem of Multiple Admissibility (MA) in language learning. Multiple Admissibility occurs in many languages when more than one grammatical form of a word fits syntactically and semantically in a given context. In second language (L2) education - in particular, in intelligent tutoring systems/computer-aided language learning (ITS/CALL) systems, which generate exercises automatically - this implies that multiple alternative answers are possible. We treat the problem as a grammaticality judgement task. We train a neural network with an objective to label sentences as grammatical or ungrammatical, using a “simulated learner corpus”: a dataset with correct text, and with artificial errors generated automatically. While MA occurs commonly in many languages, this paper focuses on learning Russian. We present a detailed classification of the types of constructions in Russian, in which MA is possible, and evaluate the model using a test set built from answers provided by the users of a running language learning system.
Text normalisation in Text-to-Speech systems is a process of converting written expressions to their spoken forms. This task is complicated because in many cases the normalised form depends on the context. Furthermore, when we analysed languages like Croatian, Lithuanian, Polish, Russian or Slovak there is additional difficulty related to their inflected nature. In this paper we want to show how to deal with this problem for one of these languages: Polish, without having a large dedicated data set and using solutions prepared for other NLP tasks. We limited our study to only numbers expressions, which are the most common non-standard words to normalise. The proposed solution is a combination of morphological tagger and transducer supported by a dictionary of numbers in their spoken forms. The data set used for evaluation is based on the part of 1-million word subset of the National Corpus of Polish. The accuracy of the described approach is presented with a comparison to a simple baseline and two commercial systems: Google Cloud Text-to-Speech and Amazon Polly.
We present experiments on Slovenian, Croatian and Serbian morphosyntactic annotation and lemmatisation between the former state-of-the-art for these three languages and one of the best performing systems at the CoNLL 2018 shared task, the Stanford NLP neural pipeline. Our experiments show significant improvements in morphosyntactic annotation, especially on categories where either semantic knowledge is needed, available through word embeddings, or where long-range dependencies have to be modelled. On the other hand, on the task of lemmatisation no improvements are obtained with the neural solution, mostly due to the heavy dependence of the task on the lookup in an external lexicon, but also due to obvious room for improvements in the Stanford NLP pipeline’s lemmatisation.