Toward Expressing Generative Lexicon Using GraphXML

Volume 2
Issue 1
Jong-Ho Lea
For machine translations and inference engines, an expressive and computationally feasible way of representing lexicons is needed.  Among many other methods in Cognitive Science, Generative Lexicon (GL) theory has gained great interests. Because GL handles four levels of lexical meaning, these combinations of four levels resolve many lexical ambiguities and explain the creative nature of language. Using only the syntactic cues, this generative mechanism resolves the ambiguities of polysemy and metonymy by finding a proper Dependency Graph between the lexical meanings.  In this paper, we propose a new way of coding the GL-style Dependency Graph of lexical information in GraphXML format.  GraphXML is a proposed XML specification to represent graphs. The graph representation can be used as a message form for lexicons as well as a data structure for inference engines.  In conjunction with the semantic graph nature, we propose a new way of representing the dynamic properties of the semantic analysis. Using C++ programming, we also implement an inference engine parsing this GraphXML form of GL.