Lexicon Generation by Extraction of Context Patterns

Semantic browser technologies such as Magpie require the construction of lexicons to support the identification of terms in Web pages which are linked to a user’s chosen ontology. We frame the generation of such lexicons from ontologies as a problem of finding synonyms and hyponyms. Synonym finding using the hypothesis of semantic substitutability relies upon the discovery of patterns in which the target word occurs. Information extraction has the potential to find a range of patterns in text. We present a methodology for finding synonyms for inclusion in lexicons in this way and preliminary tests of the method using standard tools.