A new unsupervised method for document clustering by using WordNet lexical and conceptual relations

Text document clustering provides an effective and intuitive navigation mechanism to organize a large amount of retrieval results by grouping documents in a small number of meaningful classes. Many well-known methods of text clustering make use of a long list of words as vector space which is often unsatisfactory for a couple of reasons: first, it keeps the dimensionality of the data very high, and second, it ignores important relationships between terms like synonyms or antonyms. Our unsupervised method solves both problems by using ANNIE and WordNet lexical categories and WordNet ontology in order to create a well structured document vector space whose low dimensionality allows common clustering algorithms to perform well. For the clustering step we have chosen the bisecting k-means and the Multipole tree, a modified version of the Antipole tree data structure for, respectively, their accuracy and speed.