Polyrepresentative clustering: a study of simulated user strategies and representations

Abstract

The principle of polyrepresentation and document clustering are two established methods for Interactive Information Retrieval, which have been used separately so far. In this paper we discuss a cluster based polyrepresentation approach for information need and document based representations. In our work we simulate and evaluate two possible cluster browsing strategies a user could apply to explore the polyrepresentative clusters. In our evaluation we apply information need and bibliographic features on the iSearch collection. Our results suggest that polyrepresentative cluster browsing may be more effective than exploring a ranked list.