Information-need driven query refinement

Abstract

We present a framework for query refinement that is driven by users' information needs. Since a query just approximates a user's need, we analyze the ambiguities of that query with respect to the vocabulary used for querying and the information repository. The goal is to determine the refinements that can help the user to express his original need more clearly. Particularly, a user is provided with the queries that are "nearby" the given query and a "compass" for determining the ways of changing the ambiguities of the query. The neighborhood of a query is defined regarding the result of that query. The formal concept analysis is used for its efficient calculation. We present an evaluation study that shows the benefits of our approach.