采用语义分析的标签体系构建方法

An improved method based on semantic features analysis of tag systems is proposed to solve the problem that existing tag systems are coarse-grained and their hierarchical structures are unapparent. The method analyses similarity among multiple tag systems on different websites, learns the relationships of synonym mappings and hypernym-hyponym mappings among these systems, and further obtains a fine-grained and hierarchical tag system. In order to evaluate the performance of the algorithm, two test metrics, the tag coincidence degree and the hypernym-hyponym coincidence degree, are proposed to assess the accuracy of merging and constructing tag systems. Experimental results show that the method improves the tag coincidence degree and the hypernym-hyponym coincidence degree by more than 5%, and it efficiently builds a tag system with higher precision and applicability, meanwhile, lays a foundation for establishing user profile.