A fuzzy based approach for wordsense disambiguation using morphological transformation and domain link knowledge

摘要

This paper describes a fuzzy-based methodology in order to aggregate outcomes of distinct wordsense disambiguation algorithms. The latter are derived from standard Lesk algorithm, its WorldNet extension and new interpretations of the set-intersection that accounts for various WordNet domain knowledge and part-of-speech conversion. The fuzzy preference model imitates the fuzzy Borda voting scheme. The developed algorithms are evaluated according to SenseEval 2 competition dataset, where a clear improvement to the baseline algorithm has been testified.