Towards selective user specific query expansion

Abstract

Queries submitted to a search engine are usually short and ambiguous. Query expansion is an effective way to resolve the ambiguity of a query. We attempt to make a consolidated study of the effectiveness of query logs in this regard. We report empirical findings on the effect of personalization over an anonymous query log. We have designed a novel evaluation method to quantify improvements due to personalization. In the second part of our work, we present a novel approach to intelligently detect queries whose expansions potentially lead to improvements in retrieval. Empirical results obtained using this approach clearly show significant improvements over indiscriminate query expansion.