Multilingual Word Sense Induction to Improve Web Search Result Clustering

In [12] a novel approach to Web search result clustering based on Word Sense Induction, i.e. the automatic discovery of word senses from raw text was presented; key to the proposed approach is the idea of, first, automatically inducing senses for the target query and, second, clustering the search results based on their semantic similarity to the word senses induced. In [1] we proposed an innovative Word Sense Induction method based on multilingual data; key to our approach was the idea that a multilingual context representation, where the context of the words is expanded by considering its translations in different languages, may improve the WSI results; the experiments showed a clear performance gain. In this paper we give some preliminary ideas to exploit our multilingual Word Sense Induction method to Web search result clustering.