U-M Distinguished Faculty and Graduate Student Seminar in Computation, Language, and Information Wednesday, October 24, 2001 4 PM, G115 Angell Hall, Central Campus A PROBABLISTIC ACCOUNT OF LOGICAL METONYMY Alex Lascarides University of Edinburgh (Joint work with Mirella Lapata, Saarland University) In this talk I will investigate logical metonymy, i.e., constructions involving a form of semantic type coercion, in that the semantic type of a syntactic argument of a word appears to be different from the semantic type of that argument in logical form (e.g.,``enjoy the book'' means enjoy reading the book, and ``easy problem'' means a problem that is easy to solve). The systematic variation in the interpretation of such constructions suggest a rich and complex theory of composition on the syntax/semantics interface (e.g., Pustejovsky 1995). But the generative devices which are used to model logical metonymy typically fail to exhaustively describe all the possible interpretations, or they don't rank those interpretations in terms of their likelihood. In view of this, Mirella Lapata and I have used a large corpus to acquire the meanings of metonymic verbs and adjectives, and we propose a probabilistic model which provides a ranking on the set of possible interpretations. We identify lexical semantic information automatically by exploiting the consistent correspondences between surface syntactic cues and lexical meaning. We evaluate our results against paraphrase judgements elicited experimentally from humans, and show that the model's ranking of meanings correlates reliably with human intuitions: meanings that are found highly probable by the model are also rated as plausible by the human subjects. More information about the seminar is available at http://www.si.umich.edu/~radev/cli-seminar/