Graph and Word Similarity for Word Sense Disambiguation

Abstract

This paper proposes a graph and word similarity method for word sense disambiguation. Perform word sense mapping on the context words of a given ambiguous sentence to obtain the corresponding English words; compute the English word similarity of the obtained English words, and take the English words as vertices, the semantic relations between words as edges, and the similarity values as edges to construct a disambiguation graph; the graph scoring algorithm is used to score the importance of each word sense of the disambiguated words, and the highest score is selected as the final correct word sense. The experimental results show it can effectively use English knowledge resources and obtain a disambiguation accuracy rate of 0.451 on the SemEval-2007 task#5 dataset, which can improve the accuracy of Chinese word sense disambiguation.