One sense per context cluster: Improving word sense disambiguation using web-scale phrase clustering

The performance of word sense disambiguation task is still limited by lexical context matching due to data sparse problem. In this paper we present a simple but effective method that incorporates web-scale phrase clustering results for context matching. This method is able to capture some semantic relations that are not in WordNet. Without using any additional labeled data this new approach obtained 2.11%-6.92% higher accuracy over a typical supervised classifier.