Query Aggregation in Session Search

Abstract

Session search retrieves documents for a sequence of queries in a session. Prior research demonstrated that query aggregation is an effective technique for session search. This paper proposes a novel query aggregation scheme based on the discount factor in reinforcement learning. Moreover, we compare various query aggregation schemes and investigate the best scheme for aggregating queries in session search. Evaluation conducted over TREC 2011 and 2012 shows that the proposed scheme works the best and outperforms the TREC best system as well as learned weights by learning to rank.