James Pustejovsky, Sabine Bergler
There has recently been a great deal of interest in the structure of the lexicon for natural language understanding and generation. One of the major problems encountered has been the optimal organization of the enormous amounts of lexical knowledge necessary for robust NLP systems. Modifying machine readable dictionaries into semantically organized networks, therefore, has become a major research interest. In this paper we propose a representation language for lexical information in dictionaries, and describe an interactive learning approach to this problem, making use of extensive knowledge of the domain being learned. We compare our model to existing systems designed for automatic classification of lexical knowledge.