End-to-End Sequential Metaphor Identification Inspired by Linguistic Theories

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3888–3898 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 3888 End-to-End Sequential Metaphor Identification Inspired by Linguistic Theories Rui Mao1, Chenghua Lin2, and Frank Guerin1 1Department of Computing Science, University of Aberdeen, AB24 3UE, UK 1{r03rm16, f.guerin}@abdn.ac.uk 2Department of Computer Science, University of Sheffield, S1 4DP, UK 2c.lin@sheffield.ac.uk Abstract End-to-end training with Deep Neural Net- works (DNN) is a currently popular method for metaphor identification. However, stan- dard sequence tagging models do not ex- plicitly take advantage of linguistic theories of metaphor identification. We experiment with two DNN models which are inspired by two human metaphor identification proce- dures. By testing on three public datasets, we find that our models achieve state-of-the-art performance in end-to-end metaphor identifi- cation. 1 Introduction Metaphoric expressions are common in everyday language, attracting attention from both linguists and psycho-linguists (Wilks, 1975; Glucksberg, 2003; Group, 2007; Holyoak and Stamenkovi´c, 2018). Computationally, metaphor identification is a task that detects metaphors in texts. Tradi- tional approaches, such as phrase-level metaphor identification, detect metaphors with word pairs (Tsvetkov et al., 2014; Shutova et al., 2016; Rei et al., 2017), where a target word whose metaphoricity is to be identified is given in ad- vance. However, such target words are not high- lighted in real world text data; a newer approach is sequential metaphor identification, where the metaphoricity of a target word is identified with- out knowing its position in a sentence. Therefore, it is more readily applied to support Natural Lan- guage Processing tasks. The most recent approaches (Wu et al., 2018; Gao et al., 2018) treat this as a sequence tagging task: the classified labels are only conditioned on BiLSTM (Graves and Schmidhuber, 2005) hidden states of target words. This approach is not tailor- made for metaphors; it is the same procedure to that used in other sequence tagging tasks, such as Part-of-Speech (PoS) tagging (Plank et al., 2016) and Named Entity Recognition (NER) (Lample et al., 2016). However, we have available linguis- tic theories of metaphor identification, which have not yet been exploited with Deep Neural Network (DNN) models. We hypothesise that by exploiting linguistic theories of metaphor identification in the design of a DNN architecture, the model perfor- mance can be further improved. Linguistic theories suggest that a metaphor is identified by noticing a semantic contrast between a target word and its context. This is the basis of Selectional Preference Violation (SPV) (Wilks, 1975, 1978). E.g., in the sentence my car drinks gasoline (Wilks, 1978), ‘drinks’ is identified as metaphoric, because ‘drinks’ is unusual in the con- text of ‘car’ and ‘gasoline’; a car cannot drink, nor is gasoline drinkable. Formally, a label is pre- dicted, conditioned on a target word and its con- text. An alternative approach by Group (2007) and Steen et al. (2010) is the Metaphor Identifi- cation Procedure (MIP): a metaphor is identified if the literal meaning of a word contrasts with the meaning …

Content truncated to 500 words