Table 10 left without paying the bill! a good reason to treat metonymy with conceptual graphs

Abstract

Metonymy is a valuable test bed to appreciate the adequacy of a model of referential mechanisms occurring in natural language surface forms. Our concern in this paper is to propose an alternative treatment to deal with metonymy. We benefit considerably from the ability of conceptual graphs to express semantic norms for implementing a kind of preferential semantics and make special use of its normative apparatus including the type lattice used in the application, canonical graphs, conformity relation. The originality of our approach concerns two main points:

First, we treat not only nominal metonymies, as traditional works do, but also verbal ones. Whereas verb is always, in these approaches, considered as steady, we consider that the verbal operator itself can be a metonymical reference. This not only enlarges the field covered but also allows us to reach more realistic solutions.

The second point concerns the use of classical metonymy typologies. Canned metonymies may intervene in two ways, as guides and as filters; in the first case, they render the combinatorial problem manageable by focusing the search during the resolution; in the second case, they are a means to rank the relevance of the provided solutions.