Ontology Pruning: WordNet Adaptation to Biomedical domain

This paper describes an automatic technique for adapting the WordNet noun taxonomy (Fellbaum, 1998) to the medical domain, by comparing the available senses of nouns given by WordNet with the distribution of words in the Ohsumed corpus (Hersh, Buckley, Leone and Hickam, 1994), a large selection of documents from the PubMed catalogue. Though specific in nature, our experiments are based on a very small set of assumptions about lexicosyntactic patterns in the English language, and are hence easily adapted to other domains and other languages where lexicosyntactic patterns are a reliable guide to semantic usage. This enables us in many cases to analyze ambiguous words and select the correct meaning for the biomedical domain.