User Community And Contents Co-Ranking For User Generated Content Quality Evaluation

With the popularity of mobile Internet and social networks, more and more users will generate hundreds of millions of user generated content (UGC) every day. However, due to the lower threshold of the type of users who can publish content, and any user at any time can express their ideas, which therefore, also led to the uneven quality of user generated content. But user-generated content has an important role both in the commercial value and government oversight of public opinion and so on, hence, looking for quality content in massive UGC data has become an important research topic. In this thesis, we propose a new method to assess the quality of UGC, the method uses a community of users in a social network structure, which is some kind of community structure that consists of people who have similar interests. From an intuitive point of view, the number of user communities that is attracted by content is better to represent the quality of content than the number of users. Because it shows the contents can attract more types of users and the content is more potential to become a hot spot. Therefore, we proposes UGC quality assessment algorithm based on linkage between the user community and content, referred as UCCC algorithm. UCCC algorithm uses user content network and user community network and user community-content network diagram to calculate the quality. The rationale of calculation is a mutually reinforcing relationship between user community and content. That is to say, the higher quality of user community associated with the content, the higher quality of content; and the higher quality of the content, and the higher quality of user community associated with the content.The experiments about UGC quality evaluation using real-world data have shown that UCCC outperforms competitive algorithms by a good margin in most cases than other algorithms and a user community is more useful than a single user for UGC quality evaluation.This work was supported by the National Natural Science Foundation of China," hLDA based Chinese multi-document summarization"(project approval number:61202247) and "On the management of uncertainties in Web2.0user generated content"(project approval number:71231002).