Clustering terms in the Bayesian network retrieval model:: a new approach with two term-layers

The retrieval performance of an information retrieval system usually increases when it uses the relationships among the terms contained in a given document collection. However, this creates the problem of how to obtain these relationships efficiently, and how to then use them to retrieve documents given a user's query. This paper presents a new retrieval model based on a Bayesian network that represents and exploits term relationships, overcoming these two drawbacks. An efficient learning method to capture these relationships, based on term clustering, as well as their use for retrieval purposes, is also shown. (C) 2004 Elsevier B.V. All rights reserved.