Implicit user modeling for personalized search

Abstract

Information retrieval systems (e.g., web search engines) 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, resulting in inherently non-optimal retrieval performance. For example, a tourist and a programmer may use the same word "java" to search for different information, but the current search systems would return the same results. In this paper, we study how to infer a user's interest from the user's search context and use the inferred implicit user model for personalized search . We present a decision theoretic framework and develop techniques for implicit user modeling in information retrieval. We develop an intelligent client-side web search agent (UCAIR) that can perform eager implicit feedback, e.g., query expansion based on previous queries and immediate result reranking based on clickthrough information. Experiments on web search show that our search agent can improve search accuracy over the popular Google search engine.

Contextual Search: A Computational Framework

“…This is observed by Shen et al [156] where it is reported that the performance improvement is more substantial for precision at the top 20 documents than for precision at the top 10 documents. The authors' explanation for this difference between precision at ten and precision at twenty is that after twenty browsed documents the user accumulates more interaction with the system, thus letting the system to calculate better estimations than after ten documents.…”

“…So far there is nothing personalized because the user's actions are not labeled by the user. Instead of labeling every user action, Shen et al [156] propose injecting a user's mathematical model chosen from a fixed set of user's models. The goal is to minimize a risk function of the user's actions, the system's past responses and the user's model.…”

“…Thus, the system's goal is to choose an optimal response when the last action has been observed. Shen et al [156] aim to illustrate how to construct and update a user model based on the implicit feedback information. Such an approach can naturally be modeled by classification and in particular by the minimization of a risk function.…”

“…The article by Pitkow et al [137] is a good exposition of some points of this survey. An early innovative proposal of actual prototype is by Shen et al [156]. The paper by Teevan et al [160] is not a statistical paper in a strict sense, yet it addresses some basic problems of contextual search from an implicit relevance feedback point of view.…”

Naming Functions for the Vector Space Model

“…Subsequently, we provided algorithms for computing naming functions appropriate for systems that are based on the vector space model. Somehow related work includes [4,7,1].…”