Diversity-based HITS: Web page ranking by referrer and referral diversity

Abstract

We propose a Web ranking method that considers the diversity of linked pages and linking pages. Typical link analysis algorithms such as HITS and PageRank calculate scores by the number of linking pages. However, even if the number of links is the same, there is a big difference between documents linked by pages with similar content and those linked by pages with very different content. We propose two types of link diversity, referral diversity (diversity of pages linked by the page) and referrer diversity (diversity of pages linking to the page), and use the resulting diversity scores to expand the basic HITS algorithm. The results of repeated experiments showed that the diversity-based method is more useful than the original HITS algorithm for finding useful information on the Web. © 2013 Springer International Publishing.