Data-Driven Detection of Figurative Language Use in Electronic Language Resources Wim Peters and Yorick Wilks NLP Group, Department of Computer Science University of Sheffield, U.K. Examples of figurative language use have until now been the product of linguistic intro- spection and manual lookup in dictionaries and texts. The availability of electronic se- mantic resources makes it possible to investigate the nature of regularities between sense distinctions in a data-driven way. This article describes a semiautomatic approach to se- lect instances of metonymy and metaphor from a large electronic thesaurus. The study of figurative use of language has generally been approached from a the- oretical perspective. It is mainly based on a scholar’s introspection and collection or construction of a limited number of examples, such as animal/food (lamb, chicken) and product/producer (newspaper , Honda; see Main Types of Figurative Language for a more substantial list). Most of these relations have been identified by examination of a limited quan- tity of linguistic material (texts, dictionaries) or introspection. They do not reflect the true range and distribution of the phenomenon of figurative language use. One reason for the restricted number of attested instances of metonymy and metaphor in texts is that their discovery is a time-consuming manual task. Cor- pus-oriented research into metonymy relies on the annotation of a substantial amount of texts with metonymic patterns. This has been tackled for a limited set of metonymic patterns in projects such as ACE (Automatic Content Extraction), 1 where the range of metonymy is restricted to alternations between persons, facili- ties, locations, and organizations. METAPHOR AND SYMBOL, 18(3), 161–173 Copyright © 2003, Lawrence Erlbaum Associates, Inc. Requests for reprints should be sent to Wim Peters, NLP Group, Department of Computer Science, University of Sheffield, Regent Court, 211 Portobello Street, Sheffield S1 4DP United Kingdom. E-mail: w.peters@dcs.shef.ac.uk 1 http://www.itl.nist.gov/iad/894.01/tests/ace/