| With the quick development of Web2.0applications, tag based social tagging system has become a most important information organizing method. The free way of social tagging system’s usage has greatly simplified the works need to be done during users’ information organizing process. However, on the other hand, it also leads to various problems, which limit the quality of social tagging results. To improve the quality of social tagging results, researchers have put lots of attention on tag generation methods.Based on the analysis of available social tagging system studies and tag generation methods, this thesis paper figures out that the main reason for the disability of available tag generations methods is that the density required by the algorithms of these methods could not be met in real life social tagging systems. To solve such a problem, this thesis paper proposed a relation and content fusion analysis based tag generation method. In compare with the sparse relation between resources, users and tags, resource content provides more dense relation and yet more clues for making tag generation. By extending latent Dirichlet allocation, this thesis paper model heterogeneous objects in a unified model. Meanwhile, the complete generation character of the proposed model makes it easy to calculate the relation between different objects. This thesis paper also studied the model’s parameter estimation, incremental estimation and model inference methods. Experimental results shows that the proposed methods could provide higher quality results when compared with other methods.The proposed methods is further applied in a geological survey information integration system.The system architecture, design of system models, key class design and code implementation, together with some screen captures of the running system, are proposed. By application the proposed tag generation method, more effective information integration and resource retrieval is achieved. |
PDF Full Text Request