Research work · 2001

Forming an Integrated Lexical Resource for Word Sense Disambiguation

Oi Yee Kwong

Institutional Repositories DataBase (IRDB)

Abstract

This paper reports a full-scale linkage of noun senses between two existing lexical resources, namely WordNet and Roget's Thesaurus, to form an Integrated Lexical Resource (ILR) for use in natural language processing (NLP). The linkage was founded on a structurally-based sense-mapping algorithm. About 18,000 nouns with over 30,000 senses were mapped. Although exhaustive verification is impractical, we show that it is reasonable to expect some 70-80 % accuracy of the resultant mappings. More importantly, the ILR, which contains enriched lexical information, is readily usable in many NLP tasks. We shall explore some practical use of the ILR in word sense disambiguation (WSD), as WSD notably requires a wide range of lexical information. 1

Read free full text