On the Topic Discovery Using Query Logs and Hyperlink

With the rapid growth of the World Wide Web, the amount of information in the Web has spawned on an unpredictable scale. Recently, most researchers attempt to use conventional information retrieval techniques to classify the search results. But not like the traditional document, the web page has distinct characteristics of its own. Therefore some researchers have begun to exploit the hyperlinks between Web pages. In this paper, we propose a topic discovery algorithm, which combines the query log and the hyperlink analysis. We use the query log to find the representative Web pages with respect to users’ endorsements, and combine a link-based clustering algorithm to cluster the similar topics. Each Web page is ranked according to search engine users’ endorsements and web creators’ endorsements. The experimental results show that our method performs better than the pure hyperlink analysis algorithm (ATD) in terms of topics discrimination and topic quality.