A Context Expansion Method for Supervised Word Sense Disambiguation

Feature sparseness is one of the main causes for Word Sense Disambiguation (WSD) systems to fail, as it increases the probability of incorrect predictions. In this work, we present a WSD method to overcome this problem by using an automatically-created thesaurus to append related words to a specific context, in order to improve the effectiveness of candidate selection for an ambiguous word. We treat the context as a vector of words taken from sentences, and expand it with words from the thesaurus according to their mutual relatedness. Our results suggest that the method performs disambiguation with high precision.