The current state-of-the-art test accuracy (97.42%) on the IMDB movie reviews dataset was reported by Thongtan and Phienthrakul (2019) and achieved by the logistic regression classifier trained on the Document Vectors using Cosine Similarity (DV-ngrams-cosine) proposed in their paper and the Bag-of-N-grams (BON) vectors scaled by Naïve Bayesian weights. While large pre-trained Transformer-based models have shown SOTA results across many datasets and tasks, the aforementioned model has not been surpassed by them, despite being much simpler and pre-trained on the IMDB dataset only. In this paper, we describe an error in the evaluation procedure of this model, which was found when we were trying to analyze its excellent performance on the IMDB dataset. We further show that the previously reported test accuracy of 97.42% is invalid and should be corrected to 93.68%. We also analyze the model performance with different amounts of training data (subsets of the IMDB dataset) and compare it to the Transformer-based RoBERTa model. The results show that while RoBERTa has a clear advantage for larger training sets, the DV-ngrams-cosine performs better than RoBERTa when the labeled training set is very small (10 or 20 documents). Finally, we introduce a sub-sampling scheme based on Naïve Bayesian weights for the training process of the DV-ngrams-cosine, which leads to faster training and better quality.
We propose a solution for the LSCDiscovery shared task on Lexical Semantic Change Detection in Spanish. Our approach is based on generating lexical substitutes that describe old and new senses of a given word. This approach achieves the second best result in sense loss and sense gain detection subtasks. By observing those substitutes that are specific for only one time period, one can understand which senses were obtained or lost. This allows providing more detailed information about semantic change to the user and makes our method interpretable.
In this paper we describe our solution of the LSCDiscovery shared task on Lexical Semantic Change Discovery (LSCD) in Spanish. Our solution employs a Word-in-Context (WiC) model, which is trained to determine if a particular word has the same meaning in two given contexts. We basically try to replicate the annotation of the dataset for the shared task, but replacing human annotators with a neural network. In the graded change discovery subtask, our solution has achieved the 2nd best result according to all metrics. In the main binary change detection subtask, our F1-score is 0.655 compared to 0.716 of the best submission, corresponding to the 5th place. However, in the optional sense gain detection subtask we have outperformed all other participants. During the post-evaluation experiments we compared different ways to prepare WiC data in Spanish for fine-tuning. We have found that it helps leaving only examples annotated as 1 (unrelated senses) and 4 (identical senses) rather than using 2x more examples including intermediate annotations.
The contextualized embeddings obtained from neural networks pre-trained as Language Models (LM) or Masked Language Models (MLM) are not well suitable for solving the Lexical Semantic Change Detection (LSCD) task because they are more sensitive to changes in word forms rather than word meaning, a property previously known as the word form bias or orthographic bias. Unlike many other NLP tasks, it is also not obvious how to fine-tune such models for LSCD. In order to conclude if there are any differences between senses of a particular word in two corpora, a human annotator or a system shall analyze many examples containing this word from both corpora. This makes annotation of LSCD datasets very labour-consuming. The existing LSCD datasets contain up to 100 words that are labeled according to their semantic change, which is hardly enough for fine-tuning. To solve these problems we fine-tune the XLM-R MLM as part of a gloss-based WSD system on a large WSD dataset in English. Then we employ zero-shot cross-lingual transferability of XLM-R to build the contextualized embeddings for examples in Spanish. In order to obtain the graded change score for each word, we calculate the average distance between our improved contextualized embeddings of its old and new occurrences. For the binary change detection subtask, we apply thresholding to the same scores. Our solution has shown the best results among all other participants in all subtasks except for the optional sense gain detection subtask.
While Masked Language Models (MLM) are pre-trained on massive datasets, the additional training with the MLM objective on domain or task-specific data before fine-tuning for the final task is known to improve the final performance. This is usually referred to as the domain or task adaptation step. However, unlike the initial pre-training, this step is performed for each domain or task individually and is still rather slow, requiring several GPU days compared to several GPU hours required for the final task fine-tuning. We argue that the standard MLM objective leads to inefficiency when it is used for the adaptation step because it mostly learns to predict the most frequent words, which are not necessarily related to a final task. We propose a technique for more efficient adaptation that focuses on predicting words with large weights of the Naive Bayes classifier trained for the task at hand, which are likely more relevant than the most frequent words. The proposed method provides faster adaptation and better final performance for sentiment analysis compared to the standard approach.
In this paper, we present a system for the solution of the cross-lingual and multilingual word-in-context disambiguation task. Task organizers provided monolingual data in several languages, but no cross-lingual training data were available. To address the lack of the officially provided cross-lingual training data, we decided to generate such data ourselves. We describe a simple yet effective approach based on machine translation and back translation of the lexical units to the original language used in the context of this shared task. In our experiments, we used a neural system based on the XLM-R, a pre-trained transformer-based masked language model, as a baseline. We show the effectiveness of the proposed approach as it allows to substantially improve the performance of this strong neural baseline model. In addition, in this study, we present multiple types of the XLM-R based classifier, experimenting with various ways of mixing information from the first and second occurrences of the target word in two samples.
Consulting a dictionary or a glossary is a familiar way for many humans to figure out what does a word in a particular context mean. We hypothesize that a system that can select a proper definition for a particular word occurrence can also naturally solve tasks related to word senses. To verify this hypothesis we developed a solution for the Multilingual and Cross-lingual Word-in-Context (MCL-WiC) task, that does not use any of the shared task data or other WiC data for training. Instead, it is trained to embed word definitions from English WordNet and word occurrences in English texts into the same vector space following an approach previously proposed for Word Sense Disambiguation (WSD). To estimate the similarity in meaning of two word occurrences, we compared different metrics in this shared vector space and found that L1-distance between normalized contextualized word embeddings outperforms traditionally employed cosine similarity and several other metrics. To solve the task for languages other than English, we rely on zero-shot cross-lingual transfer capabilities of the multilingual XLM-R masked language model. Despite not using MCL-WiC training data, in the shared task our approach achieves an accuracy of 89.5% on the English test set, which is only 4% less than the best system. In the multilingual subtask zero-shot cross-lingual transfer shows competitive results, that are within 2% from the best systems for Russian, French, and Arabic. In the cross-lingual subtask are within 2-4% from the best systems.
This paper presents our approaches to SemEval-2021 Task 2: Multilingual and Cross-lingual Word-in-Context Disambiguation task. The first approach attempted to reformulate the task as a question answering problem, while the second one framed it as a binary classification problem. Our best system, which is an ensemble of XLM-R based binary classifiers trained with data augmentation, is among the 3 best-performing systems for Russian, French and Arabic in the multilingual subtask. In the post-evaluation period, we experimented with batch normalization, subword pooling and target word occurrence aggregation methods, resulting in further performance improvements.
This work describes our approach for subtasks of SemEval-2021 Task 8: MeasEval: Counts and Measurements which took the official first place in the competition. To solve all subtasks we use multi-task learning in a question-answering-like manner. We also use learnable scalar weights to weight subtasks’ contribution to the final loss in multi-task training. We fine-tune LUKE to extract quantity spans and we fine-tune RoBERTa to extract everything related to found quantities, including quantities themselves.
In this paper, we describe the results of team LIORI at the FinCausal 2020 Shared task held as a part of the 1st Joint Workshop on Financial Narrative Processing and MultiLingual Financial Summarisation. The shared task consisted of two subtasks: classifying whether a sentence contains any causality and labelling phrases that indicate causes and consequences. Our team ranked 1st in the first subtask and 4th in the second one. We used Transformer-based models with joint-task learning and their ensembles.
SemEval-2020 Task 1 is devoted to detection of changes in word meaning over time. The first subtask raises a question if a particular word has acquired or lost any of its senses during the given time period. The second subtask requires estimating the change in frequencies of the word senses. We have submitted two solutions for both subtasks. The first solution performs word sense induction (WSI) first, then makes the decision based on the induced word senses. We extend the existing WSI method based on clustering of lexical substitutes generated with neural language models and adapt it to the task. The second solution exploits a well-known approach to semantic change detection, that includes building word2vec SGNS vectors, aligning them with Orthogonal Procrustes and calculating cosine distance between resulting vectors. While WSI-based solution performs better in Subtask 1, which requires binary decisions, the second solution outperforms it in Subtask 2 and obtains the 3rd best result in this subtask.
This paper describes our approach to “DeftEval: Extracting Definitions from Free Text in Textbooks” competition held as a part of Semeval 2020. The task was devoted to finding and labeling definitions in texts. DeftEval was split into three subtasks: sentence classification, sequence labeling and relation classification. Our solution ranked 5th in the first subtask and 23rd and 21st in the second and the third subtasks respectively. We applied simultaneous multi-task learning with Transformer-based models for subtasks 1 and 3 and a single BERT-based model for named entity recognition.
Out-of-vocabulary words are still a challenge in cross-lingual Natural Language Processing tasks, for which transliteration from source to target language or script is one of the solutions. In this study, we collect a personal name dataset in 445 Wikidata languages (37 scripts), train Transformer-based multilingual transliteration models on 6 high- and 4 less-resourced languages, compare them with bilingual models from (Merhav and Ash, 2018) and determine that multilingual models perform better for less-resourced languages. We discover that intrinsic evaluation, i.e comparison to a single gold standard, might not be appropriate in the task of transliteration due to its high variability. For this reason, we propose using extrinsic evaluation of transliteration via the cross-lingual named entity list search task (e.g. personal name search in contacts list). Our code and datasets are publicly available online.
Lexical substitution, i.e. generation of plausible words that can replace a particular target word in a given context, is an extremely powerful technology that can be used as a backbone of various NLP applications, including word sense induction and disambiguation, lexical relation extraction, data augmentation, etc. In this paper, we present a large-scale comparative study of lexical substitution methods employing both rather old and most recent language and masked language models (LMs and MLMs), such as context2vec, ELMo, BERT, RoBERTa, XLNet. We show that already competitive results achieved by SOTA LMs/MLMs can be further substantially improved if information about the target word is injected properly. Several existing and new target word injection methods are compared for each LM/MLM using both intrinsic evaluation on lexical substitution datasets and extrinsic evaluation on word sense induction (WSI) datasets. On two WSI datasets we obtain new SOTA results. Besides, we analyze the types of semantic relations between target words and their substitutes generated by different models or given by annotators.
Word Sense Induction (WSI) is the task of grouping of occurrences of an ambiguous word according to their meaning. In this work, we improve the approach to WSI proposed by Amrami and Goldberg (2018) based on clustering of lexical substitutes for an ambiguous word in a particular context obtained from neural language models. Namely, we propose methods for combining information from left and right context and similarity to the ambiguous word, which result in generating more accurate substitutes than the original approach. Our simple yet efficient improvement establishes a new state-of-the-art on WSI datasets for two languages. Besides, we show improvements to the original approach on a lexical substitution dataset.
We describe our solutions for semantic frame and role induction subtasks of SemEval 2019 Task 2. Our approaches got the highest scores, and the solution for the frame induction problem officially took the first place. The main contributions of this paper are related to the semantic frame induction problem. We propose a combined approach that employs two different types of vector representations: dense representations from hidden layers of a masked language model, and sparse representations based on substitutes for the target word in the context. The first one better groups synonyms, the second one is better at disambiguating homonyms. Extending the context to include nearby sentences improves the results in both cases. New Hearst-like patterns for verbs are introduced that prove to be effective for frame induction. Finally, we propose an approach to selecting the number of clusters in agglomerative clustering.
We present our system for semantic frame induction that showed the best performance in Subtask B.1 and finished as the runner-up in Subtask A of the SemEval 2019 Task 2 on unsupervised semantic frame induction (Qasem-iZadeh et al., 2019). Our approach separates this task into two independent steps: verb clustering using word and their context embeddings and role labeling by combining these embeddings with syntactical features. A simple combination of these steps shows very competitive results and can be extended to process other datasets and languages.
We present a new approach to extraction of hypernyms based on projection learning and word embeddings. In contrast to classification-based approaches, projection-based methods require no candidate hyponym-hypernym pairs. While it is natural to use both positive and negative training examples in supervised relation extraction, the impact of positive examples on hypernym prediction was not studied so far. In this paper, we show that explicit negative examples used for regularization of the model significantly improve performance compared to the state-of-the-art approach of Fu et al. (2014) on three datasets from different languages.