UGC quality evaluation based on user communities and contents

Abstract

We have stepped into the era of Web2.0. User-generated content (UGC) has become a very important component of Internet information. However, the quality of UGC varies greatly. UGC with high quality occupies only a small fraction of the entirety. We have to evaluate the quality so as to extract the good UGC with high quality. Currently most of the methods for quality evaluation integrate key words and link analysis. The newest methods add the factors of users and evaluate UGC quality with features of three aspects, i.e., users, structures and contents. But they simply consider a single user as the analysis unit. We move further and propose a new idea that the user community could reflect the UGC quality better than the single user. Hence we put forward a method for UGC quality evaluation based on user communities and their relationships with UGC contents in this paper. Experiments have been performed using real data collected from weibo that is a kind of micro-blog UGC in China. Results have shown that the proposed method is valid.