基于加权元组潜在语义分析的社会标签推荐

Social Tag Recommendation Based on Weighted Tuples Latent Semantic Analysis

社会标注系统中用户生成的标签具有随意性和弱关联性,这将导致标签推荐的精确性降低。本文基于加权元组潜在语义的三维张量结构模型,引入社会网络的结构化分析方法对相关元组进行量化加权,以构建加权的三维张量结构模型,并通过元组的潜在语义分析,得到能体现用户兴趣度的加权元组集。最后,通过典型标注网站Delicious中的用户标注数据集,验证了基于加权元组潜在语义分析的三维张量模型具有较好的标签推荐效果。

In the social tagging system, user generated tags is random and weak relevance, this will leadto reduced accuracy in tag recommendation. In this paper, based on weighted tuples latent semantic struc-ture model of three dimensional tensor, introduction of social network structured analysis method to quan-tify and weight the related tuples, to construct weighted structure model of three dimensional tensor, andthrough the latent semantic analysis of tuples, get weighted tuples sets that can reflect the measure of userinterest. Finally, Through the user marking data set of the typical Delicious websites, verified based onweighted tuples latent semantic analysis model of three dimensional tensor has better effect of tag recom-mendation.