Research On Tag Recommendation Method Based On Specific Taxonomy

With the rapid development of Web2.0users are increasingly involved in the interaction of the Internet, forming a wide application of the tag as the representative of the social markers. Around the growing number of tag applications, the performance improvement of the tag application has encountered a bottleneck. From this perspective, describing the resources and the sharing of information, the accuracy of the tag of great significance. Tag more widely, how resources are automatically generated accurate tag has become very worthy of study.On the basis of analysis of the characteristics and problems of today’s field of tag study, we propose an automatically tag recommendation method from the perspective of different tag types. This approach in line with the design of classification system-the tag classification-to judge the lack of blog tag type-the missing type tag to recommend such idea to the final research purposes. First, design tag taxonomy under the consideration of resources’ description which could be the foundation of study. Make use of regular expressions, dictionary identification, the semantic distinction and other way to achieve blog tags’ classification. In order to refine the types and themes meaning, we’ll examine the tag concept level relationships, co-occurrence distribution differences between tags and with the same layer concept dictionary, the theme of class tags is subdivided into the three-theme meaning. Followed investigate the distribution of a single type tag, as well as the joint distribution between multiple tag types to identify the relevance of existing tag types, which is given from the perspective of the two publishers and users found that a lack of blog tag type. Finally, according to blog missing type we’ll give the recommendation method based on the text similarity, the text properties itself, and the hierarchical relationship.The thesis verify the high-speed and effectiveness of the tag classification algorithm by experiment, and obtain the optimal threshold of the tag hierarchical process during experimental process. Tag recommendation method would do good to the description of resources from the quality and application value, and which will bring broader applications.