Reading between the lines : Using hierarchically structured preferences to capture query semantics

Traditionally, search engines treat the keywords entered by users as a text string, without trying to understand the sense behind the user’s need. Others, employ natural language processing techniques or external dictionaries to find what the users’ needs may be. At the same time there has been some recent work done in trying to understand the user better and model her interests to improve the relevancy of the results retrieved by the search engine. In this paper, we present a new hierarchical modeling of a user’s interests which allows for the placing of a particular preference into context and an intuitive way to order her preferences, such that higher relevancy is achieved in the retrieved results. We discuss an approach that exploits these inherent semantics from the hierarchical structure and re-ranks the retrieved results for the particular information need. We accompany our approach with a set of experimental results, which show the merits of such an approach and open new avenues of research in this area.