Making Sense of Roget
Jeremy Ellman, Robert Bell · Northumbria Research Link (Northumbria University) · 2007
Being able to resolve word senses could improve precision in Information Retrieval, machine translation or other natural language applications.However, automatic disambiguation rarely provides benefits in proportion to its costs.In this study we have tried to determine the dominant or 'most frequent' sense of an ambiguous word.By identifying this, future applications could statistically choose a 'correct' word sense the majority of the time without computation.We investigated several ambiguous nouns and the frequency of their cooccurring synonyms from Roget's thesaurus using the British National Corpus.This indicates the likelihood of the possible word senses.The results have been evaluated with some success against varying lexical resources including WordNet and the Concise Oxford Dictionary.