TEXT CATEGORIZATION EXPERIMENTS USING WIKIPEDIA

Abstract. Over the years many models had been proposed for text catego-rization. One of the most widely applied is the vector space model, assuming independence between indexing terms. Since training corpora sizes are rel-atively small – compared to ∞ – the generalization power of the learning algorithms is relatively low. Using a bigger unannotated text corpus can boost the representation and hence the learning process. Based on the work of Gabrilovich and Markovitch we use Wikipedia articles to give word dis-tributional representation for documents. Since this causes dimensionality increase, some feature clustering is needed. For this end we use LSA. 1