A Novel Approach for Conventional Web Spider Results and Log Mining
Despite of the popularity of global search engines, people still suffer from low accuracy of site search. The primary reason lies in the differences of link structures and data scale between global web and websites, which leads to failures of traditional reranking methods such as HITS, PAGE RANK and DIRECT HIT. This paper proposes a novel reranking method based on user logs with in websites. With the help of website Taxonomy, the authors mine for generalized association rules and abstract access patterns of different levels. Mining results are subsequently used to rerank the retrieved pages. One of the advantages of their mining algorithm is that it resolves the diversity problem of user's access behavior and discovers general patterns.