Log mining to improve the performance of site search

Gui-Rong Xue, Hua-Jun Zeng, Zheng Chen, Wei-Ying Ma, Chao-Jun Lu
Proceedings of the Third International Conference on Web Information Systems Engineering (Workshops), 2002.  
In despite of the popularity of current search engines, people still suffer search failure and lots of non-relevant results when finding some specific information from a specific website. This is because the site search performance is not satisfying as the whole Web search. This paper analyzes the specialty of site search compared with traditional Web search, and the non-applicability of link-based re-ranking techniques such as HITS and PageRank. In this paper, we propose to use log mining to
more » ... prove the site search performance. With the help of website taxonomy, a generalized association rule mining technique is applied to users' log to abstract the user's access patterns at different levels, and the mining results are then applied to re-ranking the retrieved pages. Our mining algorithm tackles the diversity problem of user's access behavior and mines out general patterns. The experimental results show that our proposed method outperforms keyword-based method by 15% and DirectHit by 13% respectively.
doi:10.1109/wisew.2002.1177868 dblp:conf/wise/XueZCML02 fatcat:qtzve4yix5ha7ilspb7ingmjvq