Abstract

With the development of social network services (SNS), more and more people exposure their relationships onto SNS platforms, such as Twitter and Weibo. Tag information, which is always added by users themselves, is often adopted to locate people with certain interests or characteristics. However, statistics on practical SNS data show that more than 60% users do not add any tags to themselves. The aim of our work is to predict tags for those none-tagged users so that they can be searched and grouped effectively. In this paper, we first define the problem of predicting tags for none-tagged person. Then a two-stage predicting algorithm is proposed to solve the problem. Experimental results based on a dataset with 12.8 million users of Sina weibo show effectiveness of the proposed algorithm. An SNS expert search system named Weibo Xunren (xunren.thuir.org) is also constructed based on the proposed algorithm.

Paper Details