We report the findings of the 2024 Multilingual Lexical Simplification Pipeline shared task. We released a new dataset comprising 5,927 instances of lexical complexity prediction and lexical simplification on common contexts across 10 languages, split into trial (300) and test (5,627). 10 teams participated across 2 tracks and 10 languages with 233 runs evaluated across all systems. Five teams participated in all languages for the lexical complexity prediction task and 4 teams participated in all languages for the lexical simplification task. Teams employed a range of strategies, making use of open and closed source large language models for lexical simplification, as well as feature-based approaches for lexical complexity prediction. The highest scoring team on the combined multilingual data was able to obtain a Pearson’s correlation of 0.6241 and an ACC@1@Top1 of 0.3772, both demonstrating that there is still room for improvement on two difficult sub-tasks of the lexical simplification pipeline.

We present preliminary findings on the MultiLS dataset, developed in support of the 2024 Multilingual Lexical Simplification Pipeline (MLSP) Shared Task. This dataset currently comprises of 300 instances of lexical complexity prediction and lexical simplification across 10 languages. In this paper, we (1) describe the annotation protocol in support of the contribution of future datasets and (2) present summary statistics on the existing data that we have gathered. Multilingual lexical simplification can be used to support low-ability readers to engage with otherwise difficult texts in their native, often low-resourced, languages.

With the recent advances of large language models (LLMs), it is no longer infeasible to build an automated debate system that helps people to synthesise persuasive arguments. Previous work attempted this task by integrating multiple components. In our work, we introduce an argument mining dataset that captures the end-to-end process of preparing an argumentative essay for a debate, which covers the tasks of claim and evidence identification (Task 1 ED), evidence convincingness ranking (Task 2 ECR), argumentative essay summarisation and human preference ranking (Task 3 ASR) and metric learning for automated evaluation of resulting essays, based on human feedback along argument quality dimensions (Task 4 SQE). Our dataset contains 14k examples of claims that are fully annotated with various properties supporting the aforementioned tasks. We evaluate multiple generative baselines for each of these tasks, including representative LLMs. We find, that while they show promising results on individual tasks in our benchmark, their end-to-end performance on all four tasks in succession deteriorates significantly, both in automated measures as well as in human-centred evaluation. This challenge presented by our proposed dataset motivates future research on end-to-end argument mining and summarisation. The repository of this project is available at https://github.com/HarrywillDr/ArgSum-Datatset.

The proliferation of Conversational AI agents (CAAs) has emphasised the need to distinguish between human and machine-generated texts, with implications spanning digital forensics and cybersecurity. While prior research primarily focussed on distinguishing human from machine-generated text, our study takes a more refined approach by analysing different CAAs. We construct linguistic profiles for five CAAs, aiming to identify Uniquely Identifiable Linguistic Patterns (UILPs) for each model using authorship attribution techniques. Authorship attribution (AA) is the task of identifying the author of an unknown text from a pool of known authors. Our research seeks to answer crucial questions about the existence of UILPs in CAAs, the linguistic overlap between various text types generated by these models, and the feasibility of Authorship Attribution (AA) for CAAs based on UILPs. Promisingly, we are able to attribute CAAs based on their original texts with a weighted F1-score of 96.94%. Further, we are able to attribute CAAs according to their writing style (as specified by prompts), yielding a weighted F1-score of 95.84%, which sets the baseline for this task. By employing principal component analysis (PCA), we identify the top 100 most informative linguistic features for each CAA, achieving a weighted F1-score ranging from 86.04% to 97.93%, and an overall weighted F1-score of 93.86%.

The chess domain is well-suited for creating an artificial intelligence (AI) system that mimics real-world challenges, including decision-making. Throughout the years, minimal attention has been paid to investigating insights derived from unstructured chess data sources. In this study, we examine the complicated relationships between multiple referenced moves in a chess-teaching textbook, and propose a novel method designed to encapsulate chess knowledge derived from move-action phrases. This study investigates the feasibility of using a modified sentiment analysis method as a means for evaluating chess moves based on text. Our proposed Aspect-Based Sentiment Analysis (ABSA) method represents an advancement in evaluating the sentiment associated with referenced chess moves. By extracting insights from move-action phrases, our approach aims to provide a more fine-grained and contextually aware ‘chess move’-based sentiment classification. Through empirical experiments and analysis, we evaluate the performance of our fine-tuned ABSA model, presenting results that confirm the efficiency of our approach in advancing aspect-based sentiment classification within the chess domain. This research contributes to the area of game-playing by machines and shows the practical applicability of leveraging NLP techniques to understand the context of strategic games. Keywords: Natural Language Processing, Chess, Aspect-based Sentiment Analysis (ABSA), Chess Move Evaluation.

Idiomatic expressions are used in everyday language and typically convey affect, i.e., emotion. However, very little work investigating the extent to which automated methods can recognise emotions expressed in idiom-containing text has been undertaken. This can be attributed to the lack of emotion-labelled datasets that support the development and evaluation of such methods. In this paper, we present the IDioms with EMotions (IDEM) dataset consisting of a total of 9685 idiom-containing sentences that were generated and labelled with any one of 36 emotion types, with the help of the GPT-4 generative language model. Human validation by two independent annotators showed that more than 51% of the generated sentences are ideal examples, with the annotators reaching an agreement rate of 62% measured in terms of Cohen’s Kappa coefficient. To establish baseline performance on IDEM, various transformer-based emotion recognition approaches were implemented and evaluated. Results show that a RoBERTa model fine-tuned as a sequence classifier obtains a weighted F1-score of 58.73%, when the sequence provided as input specifies the idiom contained in a given sentence, together with its definition. Since this input configuration is based on the assumption that the idiom contained in the given sentence is already known, we also sought to assess the feasibility of automatically identifying the idioms contained in IDEM sentences. To this end, a hybrid idiom identification approach combining a rule-based method and a deep learning-based model was developed, whose performance on IDEM was determined to be 84.99% in terms of F1-score.

We are currently witnessing a concerning surge in the spread of hate speech across various social media platforms, targeting individuals or groups based on their protected characteristics such as race, religion, nationality and gender. This paper focuses on the detection of hate type (Task 1) and hate target (Task 2) in the Arabic language. To comprehensively address this problem, we have combined and re-annotated hate speech tweets from existing publicly available corpora, resulting in the creation of AraTar, the first and largest Arabic corpus annotated with support for multi-label classification for both hate speech types and target detection with a high inter-annotator agreement. Additionally, we sought to determine the most effective machine learning-based approach for addressing this issue. To achieve this, we compare and evaluate different approaches, including: (1) traditional machine learning-based models, (2) deep learning-based models fed with contextual embeddings, and (3) fine-tuning language models (LMs). Our results demonstrate that fine-tuning LMs, specifically using AraBERTv0.2-twitter (base), achieved the highest performance, with a micro-averaged F1-score of 84.5% and 85.03%, and a macro-averaged F1-score of 77.46% and 73.15%, for Tasks 1 and 2, respectively.

Digital archive collections that have been contributed by communities, known as community-generated digital content (CGDC), are important sources of historical and cultural knowledge. However, CGDC items are not easily searchable due to semantic information being obscured within their textual metadata. In this paper, we investigate the extent to which state-of-the-art, general-domain entity linking (EL) models (i.e., BLINK, EPGEL and mGENRE) can map named entities mentioned in CGDC textual metadata, to Wikidata entities. We evaluate and compare their performance on an annotated dataset of CGDC textual metadata and provide some error analysis, in the way of informing future studies aimed at enriching CGDC metadata using entity linking methods.

Efforts to apply transformer-based language models (TLMs) to the problem of reasoning in natural language have enjoyed ever-increasing success in recent years. The most fundamental task in this area to which nearly all others can be reduced is that of determining satisfiability. However, from a logical point of view, satisfiability problems vary along various dimensions, which may affect TLMs’ ability to learn how to solve them. The problem instances of satisfiability in natural language can belong to different computational complexity classes depending on the language fragment in which they are expressed. Although prior research has explored the problem of natural language satisfiability, the above-mentioned point has not been discussed adequately. Hence, we investigate how problem instances from varying computational complexity classes and having different grammatical constructs impact TLMs’ ability to learn rules of inference. Furthermore, to faithfully evaluate TLMs, we conduct an empirical study to explore the distribution of satisfiability problems.

Hate speech on social media has proliferated in Ethiopia. To support studies aimed at investigating the targets and types of hate speech circulating in the Ethiopian context, we developed a new fine-grained annotation scheme that captures three elements of hate speech: the target (i.e., any groups with protected characteristics), type (i.e., the method of abuse) and nature (i.e., the style of the language used). We also developed a new lexicon of hate speech-related keywords in the four most prominent languages found on Ethiopian social media: Amharic, Afaan Oromo, English and Tigrigna. These keywords enabled us to retrieve social media posts (also in the same four languages) from three platforms (i.e., X, Telegram and Facebook), that are likely to contain hate speech. Experts in the Ethiopian context then manually annotated a sample of those retrieved posts, obtaining fair to moderate inter-annotator agreement. The resulting annotations formed the basis of a case study of which groups tend to be targeted by particular types of hate speech or by particular styles of hate speech language.

Cantonese-English is a low-resource language pair for machine translation (MT) studies, despite the vast amount of English content publicly available online and the large amount of native Cantonese speakers. Based on our previous work on CANTONMT from Hong et al. (2024), where we created the open-source fine-tuned systems for Cantonese-English Neural MT (NMT) using base-models NLLB, OpusMT, and mBART and corpus collections and creation, in this paper, we report our extended experiments on model training and comparisons. In particular, we incorporated human-based evaluations using native Cantonese speakers who are also fluent in the English language. We designed a modified version of the HOPE metric from Gladkoff and Han (2022) for the categorised error analysis and serenity-level statistics (naming HOPES). The models selected for human evaluations are NLLB-mBART fine-tuned and two translators from commercial companies: Bing and GPT4.

Neural Machine Translation (NMT) for low-resource languages remains a challenge for many NLP researchers. In this work, we deploy a standard data augmentation methodology by back-translation to a new language translation direction, i.e., Cantonese-to-English. We present the models we fine-tuned using the limited amount of real data and the synthetic data we generated using back-translation by three models: OpusMT, NLLB, and mBART.We carried out automatic evaluation using a range of different metrics including those that are lexical-based and embedding-based.Furthermore, we create a user-friendly interface for the models we included in this project, CantonMT, and make it available to facilitate Cantonese-to-English MT research. Researchers can add more models to this platform via our open-source CantonMT toolkit, available at https://github.com/kenrickkung/CantoneseTranslation.

Argument mining (AM) is a natural language processing task that aims to generate an argumentative graph given an unstructured argumentative text. An argumentative graph that consists of argumentative components and argumentative relations contains completed information of an argument and exhibits the logic of an argument. As the argument structure of an argumentative text can be regarded as an answer to a “why” question, the whole argument structure is therefore similar to the “chain of thought” concept, i.e., the sequence of ideas that lead to a specific conclusion for a given argument (Wei et al., 2022). For argumentative texts in the same specific genre, the “chain of thought” of such texts is usually similar, i.e., in a student essay, there is usually a major claim supported by several claims, and then a number of premises which are related to the claims are included (Eger et al., 2017). In this paper, we propose a new perspective which transfers the argument mining task into a multi-hop reading comprehension task, allowing the model to learn the argument structure as a “chain of thought”. We perform a comprehensive evaluation of our approach on two AM benchmarks and find that we surpass SOTA results. A detailed analysis shows that specifically the “chain of thought” information is helpful for the argument mining task.

Medical progress notes play a crucial role in documenting a patient’s hospital journey, including his or her condition, treatment plan, and any updates for healthcare providers. Automatic summarisation of a patient’s problems in the form of a “problem list” can aid stakeholders in understanding a patient’s condition, reducing workload and cognitive bias. BioNLP 2023 Shared Task 1A focusses on generating a list of diagnoses and problems from the provider’s progress notes during hospitalisation. In this paper, we introduce our proposed approach to this task, which integrates two complementary components. One component employs large language models (LLMs) for data augmentation; the other is an abstractive summarisation LLM with a novel pre-training objective for generating the patients’ problems summarised as a list. Our approach was ranked second among all submissions to the shared task. The performance of our model on the development and test datasets shows that our approach is more robust on unknown data, with an improvement of up to 3.1 points over the same size of the larger model.

This paper presents our submission to the WMT 2023 Quality Estimation (QE) shared task 1 (sentence-level subtask). We propose a straightforward training data augmentation approach aimed at improving the correlation between QE model predictions and human quality assessments. Utilising eleven data augmentation approaches and six distinct language pairs, we systematically create augmented training sets by individually applying each method to the original training set of each respective language pair. By evaluating the performance gap between the model before and after training on the augmented dataset, as measured on the development set, we assess the effectiveness of each augmentation method. Experimental results reveal that synonym replacement via the Paraphrase Database (PPDB) yields the most substantial performance boost for language pairs English-German, English-Marathi and English-Gujarati, while for the remaining language pairs, methods such as contextual word embeddings-based words insertion, back translation, and direct paraphrasing prove to be more effective. Training the model on a more diverse and larger set of samples does confer further performance improvements for certain language pairs, albeit to a marginal extent, and this phenomenon is not universally applicable. At the time of submission, we select the model trained on the augmented dataset constructed using the respective most effective method to generate predictions for the test set in each language pair, except for the English-German. Despite not being highly competitive, our system consistently surpasses the baseline performance on most language pairs and secures a third-place ranking in the English-Marathi.

Relation extraction (RE) is one of the tasks behind many relevant natural language processing (NLP) applications. Exploiting the information hidden in millions of scholarly articles by leveraging NLP, specifically RE, systems could benefit studies in specialized domains, e.g. biomedicine and biodiversity. Although deep learning (DL)-based methods have shown state-of-the-art performance in many NLP tasks including RE, DL for domain-specific RE systems has been hindered by the lack of expert-labeled datasets which are typically required to train such methods. In this paper, we take advantage of the zero-shot (i.e., not requiring any labeled data) capability of pattern-based methods for RE using a rule-based approach, combined with templates for natural language inference (NLI) transformer models. We present our hybrid method for RE that exploits the advantages of both methods, i.e., interpretability of rules and transferability of transformers. Evaluated on a corpus of biodiversity literature with annotated relations, our hybrid method demonstrated an improvement of up to 15 percentage points in recall and best performance over solely rule-based and transformer-based methods with F1-scores ranging from 89.61% to 96.75% for reproductive condition - temporal expression relations, and ranging from 85.39% to 89.90% for habitat - geographic location relations.

This paper presents the methods we developed for the Nuanced Arabic Dialect Identification (NADI) 2023 shared task, specifically targeting the two subtasks focussed on sentence-level machine translation (MT) of text written in any of four Arabic dialects (Egyptian, Emirati, Jordanian and Palestinian) to Modern Standard Arabic (MSA). Our team, UniManc, employed models based on T5: multilingual T5 (mT5), multi-task fine-tuned mT5 (mT0) and AraT5. These models were trained based on two configurations: joint model training for all regional dialects (J-R) and independent model training for every regional dialect (I-R). Based on the results of the official NADI 2023 evaluation, our I-R AraT5 model obtained an overall BLEU score of 14.76, ranking first in the Closed Dialect-to-MSA MT subtask. Moreover, in the Open Dialect-to-MSA MT subtask, our J-R AraT5 model also ranked first, obtaining an overall BLEU score of 21.10.

How do different generalised quantifiers affect the behaviour of transformer-based language models (TLMs)? The recent popularity of TLMs and the central role generalised quantifiers have traditionally played in linguistics and logic bring this question into particular focus. The current research investigating this subject has not utilised a task defined purely in a logical sense, and thus, has not captured the underlying logical significance of generalised quantifiers. Consequently, they have not answered the aforementioned question faithfully or adequately. Therefore, we investigate how different generalised quantifiers affect TLMs by employing a textual entailment problem defined in a purely logical sense, namely, model-checking with natural language. Our approach permits the automatic construction of datasets with respect to which we can assess the ability of TLMs to learn the meanings of generalised quantifiers. Our investigation reveals that TLMs generally can comprehend the logical semantics of the most common generalised quantifiers, but that distinct quantifiers influence TLMs in varying ways.

Temporal relation extraction models have thus far been hindered by a number of issues in existing temporal relation-annotated news datasets, including: (1) low inter-annotator agreement due to the lack of specificity of their annotation guidelines in terms of what counts as a temporal relation; (2) the exclusion of long-distance relations within a given document (those spanning across different paragraphs); and (3) the exclusion of events that are not centred on verbs. This paper aims to alleviate these issues by presenting a new annotation scheme that clearly defines the criteria based on which temporal relations should be annotated. Additionally, the scheme includes events even if they are not expressed as verbs (e.g., nominalised events). Furthermore, we propose a method for annotating all temporal relations—including long-distance ones—which automates the process, hence reducing time and manual effort on the part of annotators. The result is a new dataset, the TIMELINE corpus, in which improved inter-annotator agreement was obtained, in comparison with previously reported temporal relation datasets. We report the results of training and evaluating two baseline temporal relation extraction models on the new corpus, and compare them with results obtained on the widely used MATRES corpus.

This work introduces a novel three-class annotation scheme for text-based dementia classification in patients, based on their recorded visit interactions. Multiple models were developed utilising BERT, RoBERTa and DistilBERT. Two approaches were employed to improve the representation of dementia samples: oversampling the underrepresented data points in the original Pitt dataset and combining the Pitt with the Holland and Kempler datasets. The DistilBERT models trained on either an oversampled Pitt dataset or the combined dataset performed best in classifying the dementia class. Specifically, the model trained on the oversampled Pitt dataset and the one trained on the combined dataset obtained state-of-the-art performance with 98.8% overall accuracy and 98.6% macro-averaged F1-score, respectively. The models’ outputs were manually inspected through saliency highlighting, using Local Interpretable Model-agnostic Explanations (LIME), to provide a better understanding of its predictions.

Argument summarisation is a promising but currently under-explored field. Recent work has aimed to provide textual summaries in the form of concise and salient short texts, i.e., key points (KPs), in a task known as Key Point Analysis (KPA). One of the main challenges in KPA is finding high-quality key point candidates from dozens of arguments even in a small corpus. Furthermore, evaluating key points is crucial in ensuring that the automatically generated summaries are useful. Although automatic methods for evaluating summarisation have considerably advanced over the years, they mainly focus on sentence-level comparison, making it difficult to measure the quality of a summary (a set of KPs) as a whole. Aggravating this problem is the fact that human evaluation is costly and unreproducible. To address the above issues, we propose a two-step abstractive summarisation framework based on neural topic modelling with an iterative clustering procedure, to generate key points which are aligned with how humans identify key points. Our experiments show that our framework advances the state of the art in KPA, with performance improvement of up to 14 (absolute) percentage points, in terms of both ROUGE and our own proposed evaluation metrics. Furthermore, we evaluate the generated summaries using a novel set-based evaluation toolkit. Our quantitative analysis demonstrates the effectiveness of our proposed evaluation metrics in assessing the quality of generated KPs. Human evaluation further demonstrates the advantages of our approach and validates that our proposed evaluation metric is more consistent with human judgment than ROUGE scores.

British Sign Language (BSL) is a complex language with its own vocabulary and grammatical structure, separate from English. Despite its long-standing and widespread use by Deaf communities within the UK, thus far, there have been no effective tools for translating written English into BSL. This overt lack of available resources made learning the language highly inaccessible for most people, exacerbating the communication barrier between hearing and Deaf individuals. This paper introduces a rule-based translation system, designed with the ambitious aim of creating the first web application that is not only able to translate sentences in written English into a BSL video output, but can also serve as a learning aid to empower the development of BSL proficiency.

Legislative debate transcripts provide citizens with information about the activities of their elected representatives, but are difficult for people to process. We propose the novel task of policy-focused stance detection, in which both the policy proposals under debate and the position of the speakers towards those proposals are identified. We adapt a previously existing dataset to include manual annotations of policy preferences, an established schema from political science. We evaluate a range of approaches to the automatic classification of policy preferences and speech sentiment polarity, including transformer-based text representations and a multi-task learning paradigm. We find that it is possible to identify the policies under discussion using features derived from the speeches, and that incorporating motion-dependent debate modelling, previously used to classify speech sentiment, also improves performance in the classification of policy preferences. We analyse the output of the best performing system, finding that discriminating features for the task are highly domain-specific, and that speeches that address policy preferences proposed by members of the same party can be among the most difficult to predict.

Forced labour is the most common type of modern slavery, and it is increasingly gaining the attention of the research and social community. Recent studies suggest that artificial intelligence (AI) holds immense potential for augmenting anti-slavery action. However, AI tools need to be developed transparently in cooperation with different stakeholders. Such tools are contingent on the availability and access to domain-specific data, which are scarce due to the near-invisible nature of forced labour. To the best of our knowledge, this paper presents the first openly accessible English corpus annotated for multi-class and multi-label forced labour detection. The corpus consists of 989 news articles retrieved from specialised data sources and annotated according to risk indicators defined by the International Labour Organization (ILO). Each news article was annotated for two aspects: (1) indicators of forced labour as classification labels and (2) snippets of the text that justify labelling decisions. We hope that our data set can help promote research on explainability for multi-class and multi-label text classification. In this work, we explain our process for collecting the data underpinning the proposed corpus, describe our annotation guidelines and present some statistical analysis of its content. Finally, we summarise the results of baseline experiments based on different variants of the Bidirectional Encoder Representation from Transformer (BERT) model.

The goal of text zoning is to segment a text into zones (i.e., Background, Conclusion) that serve distinct functions. Argumentative zoning, a specific text zoning scheme for the scientific domain, is considered as the antecedent for argument mining by many researchers. Surprisingly, however, little work is concerned with exploiting zoning information to improve the performance of argument mining models, despite the relatedness of the two tasks. In this paper, we propose two transformer-based models to incorporate zoning information into argumentative component identification and classification tasks. One model is for the sentence-level argument mining task and the other is for the token-level task. In particular, we add the zoning labels predicted by an off-the-shelf model to the beginning of each sentence, inspired by the convention commonly used biomedical abstracts. Moreover, we employ multi-head attention to transfer the sentence-level zoning information to each token in a sentence. Based on experiment results, we find a significant improvement in F1-scores for both sentence- and token-level tasks. It is worth mentioning that these zoning labels can be obtained with high accuracy by utilising readily available automated methods. Thus, existing argument mining models can be improved by incorporating zoning information without any additional annotation cost.

Idiomatic expressions (or idioms) are phrases where the meaning of the phrase cannot be determined from the meaning of the individual words in the expression. Translating idioms between languages is therefore a challenging task. Transformer models based on contextual embeddings have advanced the state-of-the-art across many domains in the field of natural language processing. While research using transformers has advanced both idiom detection as well as idiom disambiguation, idiom translation has not seen a similar advancement. In this work, we investigate two approaches to fine-tuning a pretrained Text-to-Text Transfer Transformer (T5) model to perform idiom translation from English to German. The first approach directly translates English idiom-containing sentences to German, while the second is underpinned by idiom paraphrasing, firstly paraphrasing English idiomatic expressions to their simplified English versions before translating them to German. Results of our evaluation show that each of the approaches is able to generate adequate translations.

In this paper, we describe the approaches we developed for the Nuanced Arabic Dialect Identification (NADI) 2022 shared task, which consists of two subtasks: the identification of country-level Arabic dialects and sentiment analysis. Our team, UniManc, developed approaches to the two subtasks which are underpinned by the same model: a pre-trained MARBERT language model. For Subtask 1, we applied undersampling to create versions of the training data with a balanced distribution across classes. For Subtask 2, we further trained the original MARBERT model for the masked language modelling objective using a NADI-provided dataset of unlabelled Arabic tweets. For each of the subtasks, a MARBERT model was fine-tuned for sequence classification, using different values for hyperparameters such as seed and learning rate. This resulted in multiple model variants, which formed the basis of an ensemble model for each subtask. Based on the official NADI evaluation, our ensemble model obtained a macro-F1-score of 26.863, ranking second overall in the first subtask. In the second subtask, our ensemble model also ranked second, obtaining a macro-F1-PN score (macro-averaged F1-score over the Positive and Negative classes) of 73.544.

An in-depth analysis of the level of language understanding required by existing Machine Reading Comprehension (MRC) benchmarks can provide insight into the reading capabilities of machines. In this paper, we propose an ablation-based methodology to assess the extent to which MRC datasets evaluate the understanding of explicit discourse relations. We define seven MRC skills which require the understanding of different discourse relations. We then introduce ablation methods that verify whether these skills are required to succeed on a dataset. By observing the drop in performance of neural MRC models evaluated on the original and the modified dataset, we can measure to what degree the dataset requires these skills, in order to be understood correctly. Experiments on three large-scale datasets with the BERT-base and ALBERT-xxlarge model show that the relative changes for all skills are small (less than 6%). These results imply that most of the answered questions in the examined datasets do not require understanding the discourse structure of the text. To specifically probe for natural language understanding, there is a need to design more challenging benchmarks that can correctly evaluate the intended skills.

Debate transcripts from the UK Parliament contain information about the positions taken by politicians towards important topics, but are difficult for people to process manually. While sentiment analysis of debate speeches could facilitate understanding of the speakers’ stated opinions, datasets currently available for this task are small when compared to the benchmark corpora in other domains. We present ParlVote, a new, larger corpus of parliamentary debate speeches for use in the evaluation of sentiment analysis systems for the political domain. We also perform a number of initial experiments on this dataset, testing a variety of approaches to the classification of sentiment polarity in debate speeches. These include a linear classifier as well as a neural network trained using a transformer word embedding model (BERT), and fine-tuned on the parliamentary speeches. We find that in many scenarios, a linear classifier trained on a bag-of-words text representation achieves the best results. However, with the largest dataset, the transformer-based model combined with a neural classifier provides the best performance. We suggest that further experimentation with classification models and observations of the debate content and structure are required, and that there remains much room for improvement in parliamentary sentiment analysis.

Machine Reading Comprehension (MRC) is the task of answering a question over a paragraph of text. While neural MRC systems gain popularity and achieve noticeable performance, issues are being raised with the methodology used to establish their performance, particularly concerning the data design of gold standards that are used to evaluate them. There is but a limited understanding of the challenges present in this data, which makes it hard to draw comparisons and formulate reliable hypotheses. As a first step towards alleviating the problem, this paper proposes a unifying framework to systematically investigate the present linguistic features, required reasoning and background knowledge and factual correctness on one hand, and the presence of lexical cues as a lower bound for the requirement of understanding on the other hand. We propose a qualitative annotation schema for the first and a set of approximative metrics for the latter. In a first application of the framework, we analyse modern MRC gold standards and present our findings: the absence of features that contribute towards lexical ambiguity, the varying factual correctness of the expected answers and the presence of lexical cues, all of which potentially lower the reading comprehension complexity and quality of the evaluation data.

Debate motions (proposals) tabled in the UK Parliament contain information about the stated policy preferences of the Members of Parliament who propose them, and are key to the analysis of all subsequent speeches given in response to them. We attempt to automatically label debate motions with codes from a pre-existing coding scheme developed by political scientists for the annotation and analysis of political parties’ manifestos. We develop annotation guidelines for the task of applying these codes to debate motions at two levels of granularity and produce a dataset of manually labelled examples. We evaluate the annotation process and the reliability and utility of the labelling scheme, finding that inter-annotator agreement is comparable with that of other studies conducted on manifesto data. Moreover, we test a variety of ways of automatically labelling motions with the codes, ranging from similarity matching to neural classification methods, and evaluate them against the gold standard labels. From these experiments, we note that established supervised baselines are not always able to improve over simple lexical heuristics. At the same time, we detect a clear and evident benefit when employing BERT, a state-of-the-art deep language representation model, even in classification scenarios with over 30 different labels and limited amounts of training data.

The early phases of requirements engineering (RE) deal with a vast amount of software requirements (i.e., requirements that define characteristics of software systems), which are typically expressed in natural language. Analysing such unstructured requirements, usually obtained from users’ inputs, is considered a challenging task due to the inherent ambiguity and inconsistency of natural language. To support such a task, methods based on natural language processing (NLP) can be employed. One of the more recent advances in NLP is the use of word embeddings for capturing contextual information, which can then be applied in word analogy tasks. In this paper, we describe a new resource, i.e., embedding-based representations of semantic frames in FrameNet, which was developed to support the detection of relations between software requirements. Our embeddings, which encapsulate contextual information at the semantic frame level, were trained on a large corpus of requirements (i.e., a collection of more than three million mobile application reviews). The similarity between these frame embeddings is then used as a basis for detecting semantic relatedness between software requirements. Compared with existing resources underpinned by word-level embeddings alone, and frame embeddings built upon pre-trained vectors, our proposed frame embeddings obtained better performance against judgements of an RE expert. These encouraging results demonstrate the strong potential of the resource in supporting RE analysis tasks (e.g., traceability), which we plan to investigate as part of our future work.

We investigate changes in the meanings of words used in the UK Parliament across two different epochs. We use word embeddings to explore changes in the distribution of words of interest and uncover words that appear to have undergone semantic transformation in the intervening period, and explore different ways of obtaining target words for this purpose. We find that semantic changes are generally in line with those found in other corpora, and little evidence that parliamentary language is more static than general English. It also seems that words with senses that have been recorded in the dictionary as having fallen into disuse do not undergo semantic changes in this domain.

In this study, we propose to focus on understanding the evolution of a specific scientific concept—that of Circular Economy (CE)—by analysing how the language used in academic discussions has changed semantically. It is worth noting that the meaning and central theme of this concept has remained the same; however, we hypothesise that it has undergone semantic change by way of additional layers being added to the concept. We have shown that semantic change in language is a reflection of shifts in scientific ideas, which in turn help explain the evolution of a concept. Focusing on the CE concept, our analysis demonstrated that the change over time in the language used in academic discussions of CE is indicative of the way in which the concept evolved and expanded.

Analysis of the topics mentioned and opinions expressed in parliamentary debate motions–or proposals–is difficult for human readers, but necessary for understanding and automatic processing of the content of the subsequent speeches. We present a dataset of debate motions with pre-existing ‘policy’ labels, and investigate the utility of these labels for simultaneous topic and opinion polarity analysis. For topic detection, we apply one-versus-the-rest supervised topic classification, finding that good performance is achieved in predicting the policy topics, and that textual features derived from the debate titles associated with the motions are particularly indicative of motion topic. We then examine whether the output could also be used to determine the positions taken by proposers towards the different policies by investigating how well humans agree in interpreting the opinion polarities of the motions. Finding very high levels of agreement, we conclude that the policies used can be reliable labels for use in these tasks, and that successful topic detection can therefore provide opinion analysis of the motions ‘for free’.

The automatic analysis of emotions conveyed in social media content, e.g., tweets, has many beneficial applications. In the Philippines, one of the most disaster-prone countries in the world, such methods could potentially enable first responders to make timely decisions despite the risk of data deluge. However, recognising emotions expressed in Philippine-generated tweets, which are mostly written in Filipino, English or a mix of both, is a non-trivial task. In order to facilitate the development of natural language processing (NLP) methods that will automate such type of analysis, we have built a corpus of tweets whose predominant emotions have been manually annotated by means of crowdsourcing. Defining measures ensuring that only high-quality annotations were retained, we have produced a gold standard corpus of 1,146 emotion-labelled Filipino and English tweets. We validate the value of this manually produced resource by demonstrating that an automatic emotion-prediction method based on the use of a publicly available word-emotion association lexicon was unable to reproduce the labels assigned via crowdsourcing. While we are planning to make a few extensions to the corpus in the near future, its current version has been made publicly available in order to foster the development of emotion analysis methods based on advanced Filipino and English NLP.

Named entity recognition (NER) in social media (e.g., Twitter) is a challenging task due to the noisy nature of text. As part of our participation in the W-NUT 2016 Named Entity Recognition Shared Task, we proposed an unsupervised learning approach using deep neural networks and leverage a knowledge base (i.e., DBpedia) to bootstrap sparse entity types with weakly labelled data. To further boost the performance, we employed a more sophisticated tagging scheme and applied dropout as a regularisation technique in order to reduce overfitting. Even without hand-crafting linguistic features nor leveraging any of the W-NUT-provided gazetteers, we obtained robust performance with our approach, which ranked third amongst all shared task participants according to the official evaluation on a gold standard named entity-annotated corpus of 3,856 tweets.

The process of annotating text corpora involves establishing annotation schemata which define the scope and depth of an annotation task at hand. We demonstrate this activity in Argo, a Web-based workbench for the analysis of textual resources, which facilitates both automatic and manual annotation. Annotation tasks in the workbench are defined by building workflows consisting of a selection of available elementary analytics developed in compliance with the Unstructured Information Management Architecture specification. The architecture accommodates complex annotation types that may define primitive as well as referential attributes. Argo aids the development of custom annotation schemata and supports their interoperability by featuring a schema editor and specialised analytics for schemata alignment. The schema editor is a self-contained graphical user interface for defining annotation types. Multiple heterogeneous schemata can be aligned by including one of two type mapping analytics currently offered in Argo. One is based on a simple mapping syntax and, although limited in functionality, covers most common use cases. The other utilises a well established graph query language, SPARQL, and is superior to other state-of-the-art solutions in terms of expressiveness. We argue that the customisation of annotation schemata does not need to compromise their interoperability.