Extending the Rocchio Relevance Feedback Algorithm to Provide Contextual Retrieval

摘要

Contextual retrieval supports differences amongst users in their information seeking requests. The Web, which is very dynamic and nearly universally accessible, is an environment in which it is increasingly difficult for users to find documents that satisfy their specific information needs. This problem is amplified as users tend to use short queries. Contextual retrieval attempts to address this problem by incorporating knowledge about the user and past retrieval results in the search process. In this paper we explore a feedback technique based on the Rocchio algorithm that significantly reduces demands on the user while maintaining comparable performance on the Reuters-21578 corpus.