Combining Lexical Stability and Improved Lexical Chain for Unsupervised Word Sense Disambiguation

Word Sense Disambiguation (WSD) is a traditional AI-hard problem. An improvement of WSD would have a significant impact on applications such as knowledge acquisition, text mining, information extraction, etc. Lexical chain holds a set of semantically related words of a text and provides an effective way for WSD, but existing lexical chain systems have inaccuracies in WSD for lacking a weighting scheme to measure the weights of word senses. In this paper, we propose a new unsupervised WSD method based on lexical stability and improved lexical chain. This method can disambiguate all words with a high accuracy. We evaluate the performance of our algorithm on SemCor corpus which is widely used for evaluating the accuracy of the WSD algorithm. Empirical results show that the algorithm can achieve a significant higher accuracy than state-of-the-art result.