Abstract:

With  the increase of activity on the Internet from people, the social contact based on the Internet closes that in the real world. We can structure a real social network via the search technology from the Internet. Social network search technology has captured the attention of many researchers in recent times. When multiple persons share the same name, it is essential for social network search to disambiguate them on each Web. A character weight calculation method based on Cvalue and IDF is presented so that we can retrieve accurate characters and reduce vector dimension. An algorithm based on the cosine angle is given to calculate the degree of similarity. By analyzing hierarchical and partitioning clustering in the text clustering algorithm, an improved hierarchical method to implement identical name judgment is proposed. For reducing the time complexity of clustering algorithm, a new method on calculating the centroid of cluster is presented. We test the method on a search engine for name search, and the results show that identical name judgment based on a modified hierarchical clustering algorithm can significantly improve performance.

Key words: social network;vector space model;identical judgment;hierarchical clustering