Verb Metaphor Detection via Contextual Relation Learning

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 4240–4251 August 1–6, 2021. ©2021 Association for Computational Linguistics 4240 Verb Metaphor Detection via Contextual Relation Learning Wei Song1∗, Shuhui Zhou1∗, Ruiji Fu2,3, Ting Liu4, Lizhen Liu1 1College of Information Engineering and Academy for Multidisciplinary Studies, Capital Normal University, Beijing, China 2State Key Laboratory of Cognitive Intelligence, iFLYTEK Research, China 3iFLYTEK AI Research (Hebei), Langfang, China 4Research Center for Social Computing and Information Retrieval, Harbin Institute of Technology, Harbin, China {wsong, shzhou, liz_liu7480}@cnu.edu.cn, rjfu@iflytek.com, tliu@ir.hit.edu.cn Abstract Correct natural language understanding re- quires computers to distinguish the literal and metaphorical senses of a word. Recent neu- ral models achieve progress on verb metaphor detection by viewing it as sequence labeling. In this paper, we argue that it is appropri- ate to view this task as relation classifica- tion between a verb and its various contexts. We propose the Metaphor-relation BERT (Mr- BERT) model, which explicitly models the re- lation between a verb and its grammatical, sen- tential and semantic contexts. We evaluate our method on the VUA, MOH-X and TroFi datasets. Our method gets competitive results compared with state-of-the-art approaches. 1 Introduction Metaphor is ubiquitous in our daily life for effec- tive communication (Lakoff and Johnson, 1980). Metaphor processing has become an active research topic in natural language processing due to its im- portance in understanding implied meanings. This task is challenging, requiring contextual semantic representation and reasoning. Various contexts and linguistic representation techniques have been explored in previous work. Early methods focused on analyzing restricted forms of linguistic context, such as subject- verb-object type grammatical relations, based on hand-crafted features (Shutova and Teufel, 2010b; Tsvetkov et al., 2013; Gutiérrez et al., 2016). Later, word embeddings and neural networks were in- troduced to alleviate the burden of feature engi- neering for relation-level metaphor detections (Rei et al., 2017; Mao et al., 2018). However, although grammatical relations provide the most direct clues, other contexts in running text are mostly ignored. Recently, token-level neural metaphor detection draws more attention. Several approaches discov- ∗These authors contributed equally to this work. ered that wider context can lead to better perfor- mance. Do Dinh and Gurevych (2016) considered a fixed window surrounding each target token as context. Gao et al. (2018) and Mao et al. (2018) argued that the full sentential context can provide strong clues for more accurate prediction. Some recent work also attempted to design models moti- vated by metaphor theories (Mao et al., 2019; Choi et al., 2021). Despite the progress of exploiting sentential con- text, there are still issues to be addressed. First of all, a word’s local context, its sentential con- text and other contexts should be all important for detecting metaphors; however, they are not well combined in previous work. More importantly, as shown in Figure 1, most token-level metaphor de- tection methods formulate metaphor detection as either a single-word classification or a sequence labeling problem (Gao …

Content truncated to 500 words