Information-need driven query refinement

Abstract

In this paper we presented a framework for query refinement which is driven by user's information needs. Based on the analyses of the real web IR case studies we model the query refinement process as the process of decreasing query ambiguity with respect to the user's need. In the refinement process a user is provided with the query's neighborhood and the guidelines for changing the ambiguities of the query. The approach assumes the existence of a controlled vocabulary, which is used for describing the content of information resources. We model the queries (query-answering pairs), ambiguities and refinements formally. The partial ordering of the query-answering pairs in the searching space represents the mathematical lattice structure, so that formal concept analysis is used as the mathematical backbone of the query refinement process. We present a case study related to the searching in MEDLINE, which shows the benefits of using our approach.