Enhanced Metaphor Detection via Incorporation of External Knowledge Based on Linguistic Theories

Abstract

Use of external knowledge is an important and effective method applied widely in metaphor detection. Although existing knowledge-based methods perform well, when leveraging external knowledge, they take little consideration on linguistic theories of metaphor detection. Based on Metaphor Identification Procedure (MIP) and Select Preference Violation (SPV), directly using examples and definitions of words from the Oxford Dictionary, we propose two BERT-based models for metaphor detection: ExampleBERT and DefinitionBERT. Experimental results show that our methods achieve state-of-the-art performance on two established metaphor datasets. Furthermore, we show that our DefinitionBERT is highly interpretable.