CSAIL Event Calendar: Previous Series
Probabilistic Dimension Reduction Methods for Information Retrieval and Web Mining
Speaker: Thomas Hofmann , Brown University
Many problems in information retrieval, natural language processing, and text mining involve estimating probabilities in very large discrete state spaces. SVD-based dimension reduction techniques like Latent Semantic Analysis as well as clustering techniques in various flavors have been popular choices to deal with the problem of data sparseness. This talk will present a probabilistic view on dimension reduction techniques which links geometrical concepts with ideas and methods from latent class modeling. Experimental results from ad hoc retrieval, collaborative filtering, language modeling, text categorization, and Web analysis are presented to stress the broad applicability and usefulness of our approach.