Web Appearance Disambiguation of Personal Names Based on Network Motif

Searching for information about a particular person is a common activity on search engines. However, current search engines do not provide any special function for search a person. Previous research has solved the problem by using additional background knowledge, such as a friend list, to cluster the searched Web pages. However, it is still difficult to retrieve and choose suitable background knowledge. In this paper, we propose a Web appearance disambiguation (WAD) system to solve the problem by only using the hyperlink structures between Web pages. The key idea of the WAD system is to find out smaller node motifs as evidences of close relationship between pages for clustering searched Web pages. Our experimental results show that, under no background knowledge, the performance of the WAD system achieves 70% for the F-measure