Collaborative tagging systems and social networks, which are all representative Web2.0 applications, have been calling attentions from both the industry and the academy. Both the fields consider tag as a user contributed classifying method and believe that tags could be efficiently used to improve information retrieval from Web and attract more people to join in the information sharing. Nevertheless, the complex nature of tagging data brings the research area with great challenges. After the first round of preliminary research work, ranging from data analysis, semantic structure mining, to the application of tags with Web page recommendation and ranking, the researchers found that more work is necessary for a deep understanding of user tagging activity. More over, the application of tagging data remains at a preliminary level due to the lake of insightful understanding of data. For the social network area, the research work from social science and physics have set up solid research background and frameworks. But, their methodologies can not directly answer the questions that how to define the mechanism of social network construction from users' open online activities and how to make the user connections improve knowledge diffusion. Basing on the existing works in the area of collaborative tagging and social network, we try to answer these questions with three aspects: 1. We analyze the data of social collaborative tagging and social network, particularly new measure is proposed on the tri-partite graph of tagging data. The statistical result reveals the potential underlying principles of user activities and its usefulness for further application. 2. We adopt tagging information to Web page recommendation methods from two aspects: tags as features and as user navigation terminals. A tri-partite graph based user navigation model is proposed, which works as a framework for random-walk based similarity calculation to involve tagging activity into recommendation. This framework is flexible in that it can shift in its structure to model different user navigation types. 3. We evaluate the diffusion of tags in the social network using the model from percolation and complex network theory. According to the diffusion ability, we define importance measures in social network to identify influential users. Further more, we initially introduce social network into graph model based Web page recommendation to evaluate the efficiency of user networks to improve information discovery.
修改评论