An Empirical Study of Learning-Based Web Search

Although there are various approaches to facilitate the information search on the Web, most current Web search and query systems only return URLs of relevant pages. Learning-based Web search is invented targeting at processing the URLs to dig out the desired information by utilizing user feedback. However, the involvement of user behavior makes the study of system performance rather complex. In this paper, we introduce the empirical study of a learning-based Web query processing system, named FACT. Four major aspects of user behavior, namely, selection rule, training strategy, training size and training iteration, are considered to show their effects on the learning results. The experimental results are presented, together with analysis for the relationships between user behavior and system performance, which are important for further improvement on learning-based Web search technology.