Towards a Corpus Annotated for Metonymies: the Case of Location Names

At the moment, language resources do not contain the necessary information for large-scale metonymy processing. As a contribution, we here present a corpus annotated for metonymies. We describe a framework for annotating metonymies in domain-independent text that considers the regularity, productivity and underspecification of metonymic usage. We then present a fully worked out annotation scheme for location names and a gold standard corpus containing 2000 annotated location names. The annotation scheme is rigorously evaluated as to its reliability and compared to previous metonymy classification proposals. In particular, we show that it is not sufficient to rely on intuitions for reliable metonymy identification and that an annotation effort with trained annotators and explicit guidelines is necessary.