Exploiting Personal Search History to Improve Search Accuracy

Personal search history is an important type of personal information, from which we can learn a user’s interests and information needs, thus improve the search service for the user. In this paper, we describe our recent work on User-Centered Adaptive Information Retrieval (UCAIR), which aims at capturing personal search history with a client-side search agent and exploiting the history information to help a user optimize search results. We propose a decision theoretic framework and develop techniques for implicit user modeling based on a user’s personal search history. We propose several context-sensitive retrieval algorithms based on statistical language models to combine the personal search history with the current query for better ranking of documents. Using these techniques, we have developed an intelligent client-side web search agent, i.e., the UCAIR search agent, which can automatically capture a user’s personal search history, store it on the local disk, and exploit it to provide personalized search.