This research develops a new approach to the lexicon, using network structures to describe the relationships between the alternative meanings of words. In traditional theories of the lexicon, these various meanings are simply listed, as though there were no further lexical structure present. Recent linguistic research suggests, however, that the lexicon is very highly structured, in ways not modeled by traditional approaches. This investigation clarifies lexical network structure in much more detail than has previously been done, and models these results in ways that will be useful to artificial intelligence researchers building computer systems for natural language processing. The lexicon - the body of knowledge about words, their meanings, and how they are related - has recently become a focus for intense research for natural language processing systems. The present project, a collaboration between researchers in artificial intelligence and linguistics, offers a new approach to capturing knowledge about words. The network approach is meant to be compact (due to a limited number of links allowed between meanings), yet powerful enough to substantially enhance the ability of computers to process natural language input and output.
