Microblog Social Relationship Mining

Since the beginning of the21st century, the appearance of Web2.0spawns the development of social networks. As the real world, the virtual communities also contain many complex social relationships. Traditionally, the research on the social relationships focuses on friend recommendation, friend rank and friend group mining, etc. However, few research has been taken on the social relationships itself. Here, two kinds of social relationships have been studied. The one is the social relationship based on topics, which belongs to the relationship between persons and things and treats things as the centre. This social relationship contains three sub-relationships:support relationships, oppose relationships and neutral relationships. The other is the social relationship based on users, which belongs to the relationship between persons and persons and treats persons as the centre. And this contains six sub-relationships:support relationships, oppose relationships, neutral relationships, alliance relationships, hostile relationships and cooperation relationships.①Microblog topics social relationships. In this study, the analysis of microblog topics social relationships is treated as the problem of short text classification. Three feature selection methods are selected, including:human selection, information gain selection and fusion selection. Tested on the dataset, the results show that the fusion selection method outperforms the other two method.②Microblog users social relationships. In this study, the analysis of microblog users social relationships is treated as a long and short text classification problem, which can be solved by using the text classification methods. Two analysis methods are used, including:single analysis and overall analysis. Second classification algorithm is adopted during the overall analysis. Finally, the experiments demonstrate that the results of second classification method is better than those of the SVM-based method.All the proposed methods have been successfully applied to two systems, including mircoblog social relationship system and haitian social relationship system. The mircoblog system is deployed in sina official applications and mainly analyzes mircoblog users’social relationship. And the haitian system is deployed in our laboratory’s official website and displays the social relationship, which can be diffused by topic social relationship and user social relationship.