Web Page Scoring Based on Link Analysis ofWeb Page Sets

We propose a new Web page scoring method based on the link analysis among sets of Web pages. Conventional link analyses such as PageRank and HITS calculate importance degree of each Web page; however, the authors of Web pages often create multiple pages to describe a specific topic. The importance degrees of such multiple Web pages cannot be derived by the conventional link analyses accurately. To cope with this problem, we need to treat the Web pages with the same contents edited by the same author as a Web page set (WPS). After constructing the link structure among WPSs, we calculate their importance degrees by using conventional link analysis schemes. In this paper, we compared our approach with the conventional method by using the NTCIR test collection, and found that our approach was better than the conventional method in terms of both WRR and DCG evaluation measures.