User-Centered Adaptive Information Retrieval
- User-Centered Adaptive Information Retrieval
- Shen, Xuehua
- Issue Date
- information retrieval
- Information retrieval systems are critical for overcoming information overload. A major deficiency of existing retrieval systems is that they generally lack user modeling and are not adaptive to individual users; information about the actual user and search context is largely ignored. Personalization is expected to break this deficiency and significantly improve retrieval accuracy. In this thesis, we study how to put the user in the center of information retrieval process for the personalized search. We develop a decision-theoretic framework for optimizing interactive information retrieval based on eager user model updating, in which the system responds to every action of the user by choosing a system action to optimize a utility function. The framework emphasizes immediate and frequent feedback to bring maximum benefit of context to the user. In general, it serves as a roadmap for studying retrieval models for personalized search. Using the general decision-theoretic framework described above, specific retrieval models for exploiting implicit user context based on statistical language model are developed to improve retrieval accuracy. Evaluation indicates that the user context information especially the clickthrough information can effectively and efficiently improve retrieval performance with no additional effort from the user. Sometimes we need user effort to provide more information to improve the retrieval performance. In this scenario, we study how a retrieval system can perform active feedback, i.e., how to choose documents for relevance feedback so that the system can learn most from the feedback information. We frame the problem of active relevance feedback as a statistical decision problem, and examine several special cases in refining the framework. We derive several practical algorithms for active feedback. The experimental results indicate that the diversity in the presented documents is a desirable property. On the result representation side, we study how to exploit a user's clickthrough information to adaptively reorganize the clustering results and help a user find the relevant information more quickly. We propose four strategies for adapting clustering results based on user actions. The simulation experiments show that the adaptation strategies have different performance for different types of users. We also conduct a user study on one of the four adaptive clustering strategies to see if an adaptive clustering system using such a strategy can bring users better search utility than a static clustering system. The results show that there is generally no significant difference between the two systems from a user's perspective. We design and develop a client-side web search agent UCAIR on top of popular search engines for personalized search. UCAIR search agent captures and exploits implicit context information such as related immediately preceding query and viewed document summaries to immediately rerank any documents that have not yet been seen by the user. User studies show that the UCAIR search agent improves performance over a popular search engine, on which UCAIR search agent is built.
- Type of Resource
- Copyright and License Information
- You are granted permission for the non-commercial reproduction, distribution, display, and performance of this technical report in any format, BUT this permission is only for a period of 45 (forty-five) days from the most recent time that you verified that this technical report is still available from the University of Illinois at Urbana-Champaign Computer Science Department under terms that include this permission. All other rights are reserved by the author(s).
Edit Collection Membership