NER4ID at SemEval-2022 Task 2: Named Entity Recognition for Idiomaticity Detection

Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 204 - 210 July 14-15, 2022 ©2022 Association for Computational Linguistics NER4ID at SemEval-2022 Task 2: Named Entity Recognition for Idiomaticity Detection Simone Tedeschi Babelscape & Sapienza University of Rome tedeschi@babelscape.com Roberto Navigli Sapienza University of Rome navigli@diag.uniroma1.it Abstract Idioms are lexically-complex phrases whose meaning cannot be derived by compositionally interpreting their components. Although the automatic identification and understanding of idioms is essential for a wide range of Nat- ural Language Understanding tasks, they are still largely under-investigated. This motivated the organization of the SemEval-2022 Task 2, which is divided into two multilingual sub- tasks: one about idiomaticity detection, and the other about sentence embeddings. In this work, we focus on the first subtask and propose a Transformer-based dual-encoder architecture to compute the semantic similarity between a potentially-idiomatic expression and its con- text and, based on this, predict idiomaticity. Then, we show how and to what extent Named Entity Recognition can be exploited to reduce the degree of confusion of idiom identification systems and, therefore, improve performance. Our model achieves 92.1 F1 in the one-shot setting and shows strong robustness towards unseen idioms achieving 77.4 F1 in the zero- shot setting. We release our code at https: //github.com/Babelscape/ner4id. 1 Introduction One of the main challenges in Natural Language Processing (NLP) is to embed the meaning of a piece of raw text (e.g. a word or a sentence) in a low-dimensional dense vector. With the advent of pretrained language models, which exploit con- textual information and assume compositionality of word representations, significant improvements have been made in this direction (Peters et al., 2018; Devlin et al., 2019). On the other hand, very little attention has been paid to idiomatic expressions, i.e. multi-word expressions (MWEs) with an es- tablished meaning unrelated to the meanings of the individual constituents. However, since idiomatic- ity is a frequent phenomenon that can be observed in all languages, idiomatic expressions should play an important role in NLP. Indeed, their identifica- tion and understanding is crucial not only for Nat- ural Language Understanding tasks such as Word Sense Disambiguation (Bevilacqua et al., 2021b), Semantic Role Labeling (Conia et al., 2021) and Se- mantic Parsing (Bevilacqua et al., 2021a), but also for Machine Translation (Edunov et al., 2018; Liu et al., 2020), Question Answering (Mishra and Jain, 2016) and Text Summarization (Chu and Wang, 2018), inter alia. In the SemEval-2022 Task 2: Multilingual Id- iomaticity Detection and Sentence Embedding (Tayyar Madabushi et al., 2022), research on id- ioms has been promoted by adapting datasets and tasks from the work carried out by Tayyar Mad- abushi et al. (2021). Specifically, the organizers propose two subtasks: • Subtask A: a binary classification task in which potentially-idiomatic expressions (PIEs) must be labeled as either "Idiomatic" or "Literal", based on the context they appear in. To better test models’ generalization ca- pabilities, two different settings are provided: zero-shot and one-shot; • Subtask B: requires models to output the cor- rect Semantic Text …

Content truncated to 500 words