A conceptual semantics ontology for conceptual graphs

Abstract

The conceptual graph formalism in essence just specifies the form of representation. No real claim is made about the content of representation, although there is a preference for the cases underlying syntactic relations. These cases are taken from Fillmore's Case Theory that basically enforces a one-to-one correspondence between syntactic structure and conceptual primitives. But the conceptual graph of a sentence need not be similar to its syntactic structure.

In this paper we will investigate a theory of conceptual structure that is in principle, unrestricted by syntax. Jackendoff proposes Conceptual Semantics that is assumed to give a primitive deep-level semantics. The concepts and relations are based on their support for encoding visual (or other sensory) units, as well as on a preference for syntactic occurrance.

As has been remarked before, the Conceptual Semantics structures can be easily formulated as Conceptual Graphs. The result of this translation is a set of concept and relation types together with their canonical graphs. This canonical basis should be viewed as a Conceptual Semantics ontology for Conceptual Graphs.

The claim that the analyses are at a deep and primitive level is taken to mean that any commonality in meaning of words must be shown by a common underlying conceptual graph. Moreover the types used may not have any commonality themselves, and the preference for grammaticality may be violated in order to achieve primitiveness. A consequence of this is that the traditional thematic roles and cases are not primitives of the ontology, but rather relational notions that should be defined in terms of conceptual substructures.