Research Of Social Expert Recommendation In Web2.0 Environments

With the rapid development of Internet and the popularization of Web 2.0 applications, a wide variety of virtual communities get the favor of more and more people including online support forums, Q&A community and social networking sites. The virtual community has become an important channel for people to get questions and share knowledge. Meanwhile, with the "explosive" growth of Web information resources, the virtual community has accumulated huge amounts of data, but also led to the "information overload" problem getting worse. Although people have more opportunities to get their required information and knowledge in the case of abundant information resources, but also brings greater challenges, people had to endure to obtain objective information from huge amounts of data in a heavy burden. Thus, both quickly and accurately finding the information or knowledge according to the user's needs, and effectively recommending to the user, which is an important and challenging issue.Social expert recommendation plays an important role in helping users acquire effective solutions in time, as well as improving knowledge sharing in communities in Web 2.0 environments. The purposes of social expert recommendation are discovering these users with professional expertise in virtual community, and recommending it to those who need the information. In this way, users can quickly get help from experts and share their tacit knowledge,which is much helpful for solving the problem of information overload, improving user experience and level of knowledge sharing. In recent years, social expert recommendation has aroused widespread concern of academia and industry.However, the existed researches on expert recommendation haven't solved the problem of low recommendation precision in the web2.0 environment caused by data sparsity, information quality and cheating behaviors because of not taking advantage of multiple expert evidences effectively. What's more, sometimes the recommended experts cannot respond to users without delay as user context information is seldom taken into consideration in expert recommendation processes.Based on the characteristics of the virtual community, this research will propose a user-expertise mixed graph and organically integrate multiple evidences and context information into the graph, meanwhile study social expert recommendation technology based on the evidence-enhanced mixed graph to solve the problems mentioned above. The main research contents are as follows.1) Elaborating the theories and techniques related to the social expert recommendation, including the traditional theory and methods of expert recommendation, the theory of social expert recommendation under the Web2.0 environment, commonly used method of social network analysis, recommendation algorithm based on mixed graphs, and so on. The above research is the solid foundation of this paper.2) Proposing a generation method of the user-expertise bipartite graph based on the extended LDA topic model, and generating a user-expertise enhanced mixed graph on the basis of integrating multiple evidences such as supporting documents, social networks, user evaluations and context information into the bipartite graph.3) Presenting a two-stage generation method of expert recommendation, including the expert recommendation algorithm using random walk with restart and the context-based pre- and post-filtering method for social expert recommendation.