Automatic query expansion based on directed divergence

To extract informative terms through known relevance information for query expansion is an important issue in the design of an effective text retrieval system. Various discriminant functions have been developed as a device for eliciting good terms. In this paper the notion of informativeness of terms is interpreted and discriminant functions are formalised both based on information-theoretic ideas for automatic query expansion. Preliminary results show that the approach adopted in this paper is worthy of continued investigation.