Discretization Based Learning for Information Retrieval

Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP), pages 153–160, Vancouver, October 2005. c⃝2005 Association for Computational Linguistics Discretization Based Learning Approach to Information Retrieval Dmitri Roussinov Weiguo Fan Department of Information Systems W.P. Carey School of Business Department of Information Systems and Computer Science Arizona State University Virginia Polytechnic Institute and State University Tempe, AZ, 85287 Blacksburg, VA 24061 dmitri.roussinov@asu.edu wfan@vt.edu Abstract We approached the problem as learning how to order documents by estimated relevance with respect to a user query. Our support vector machines based classifier learns from the relevance judgments available with the standard test collections and generalizes to new, previously unseen queries. For this, we have designed a representation scheme, which is based on the discrete representation of the local (lw) and global (gw) weighting functions, thus is capable of reproducing and enhancing the properties of such popular ranking functions as tf.idf, BM25 or those based on language models. Our tests with the standard test collections have demonstrated the capability of our approach to achieve the performance of the best known scoring functions solely from the labeled examples and without taking advantage of knowing those functions or their important properties or parameters. 1. Introduction Our work is motivated by the objective to bring closer numerous achievements in the domains of machine learning and classification to the classical task of ad-hoc information retrieval (IR), which is ordering documents by the estimated degree of relevance to a given query. Although used with striking success for text categorization, classification-based approaches (e.g. those based on support vector machines, Joachims, 2001 ) have been relatively abandoned when trying to improve ad hoc retrieval in favor of empirical (e.g. vector space, Salton & McGill, 1983) or generative (e.g. language models; Zhai & Lafferty 2001; Song & Croft; 1999), which produce a ranking function that gives each document a score, rather than trying to learn a classifier that would help to discriminate between relevant and irrelevant documents and order them accordingly. A generative model needs to make assumptions that the query and document words are sampled from the same underlying distributions and that the distributions have certain forms, which entail specific smoothing techniques (e.g. popular Dirichlet-prior). A discriminative (classifier-based) model, on the other side, does not need to make any assumptions about the forms of the underlying distributions or the criteria for the relevance but instead, learns to predict to which class a certain pattern (document) belongs to based on the labeled training examples. Thus, an important advantage of a discriminative approach for the information retrieval task, is its ability to explicitly utilize the relevance judgments existing with standard test collections in order to train the IR algorithms and possibly enhance retrieval accuracy for the new (unseen) queries. Cohen, Shapire and Singer (1999) noted the differences between ordering and classification and presented a two-stage model to learn ordering. The first stage learns a classifier for preference relations between objects using any suitable learning mechanism (e.g. …

Content truncated to 500 words