Collecting Topic-Related Web Pages for Link Structure Analysis by Using a Potential Hub and Authority First Approach

摘要

Constructing a base set consisting of topic-related web pages is a preliminary step for those web mining algorithms which use the link structure analysis technique based on HITS. However, except checking the anchor text of links and the content of pages, there has been few of research addressing other possibilities to improve topic relevance while collecting the base set. In this paper, we propose a potential hub and authority first (PHA-first) approach utilizing the concept of hub and authority to filter web pages. We investigate the satisfaction of dozens of users about the pages recommended by our method and HITS on different topics. The results indicate that our method is superior to HITS in most cases. In addition, we also evaluate the recall and precision measures of our method. The results show that our method is with relative high precision and low recall for all topics.