William J. Rapaport and Michael W. Kibby
This project concerns the development and implementation of a computational theory of how human readers and other natural-language-understanding systems can automatically expand their vocabulary by determining the meaning of a word from context. The word might be unknown to the reader, familiar but misunderstood, or familiar but being used in a new sense. `Context' includes the prior and immediately surrounding text, grammatical information, and the reader's background knowledge, but no access to a dictionary or other external source of information (including a human). The fundamental thesis is that the meaning of such a word (1) can be determined from context, (2) can be revised and refined upon further encounters with the word, (3) "converges" to a dictionary-like definition if enough context has been provided and there have been enough exposures to the word, and (4) eventually "settles down" to a "steady state", which, however, is always subject to revision upon further encounters with the word. The system is being implemented in SNePS, which provides a software laboratory for testing and experimenting with the theory.
For further details, go to the CVA homepage.