Metaphor Interpretation Using Word Embeddings

We suggest a model for metaphor interpretation using word embeddings trained over relatively large corpus. Our system handles nominal metaphors, like time is money. It generates ranked list of potential interpretations given metaphors. Candidate meanings are drawn from collocations the topic (time) and vehicle (money) components, automatically extracted dependency-parsed explore adding candidates derived association norms (common human responses to cues). ranking procedure considers similarity between candidate measured in semantic vector space. Lastly, clustering algorithm removes semantically related duplicates, thereby allowing other attain higher rank. evaluate different sets annotated with encouraging preliminary results.