基于模块化聚类的标签弹性推荐

摘要

[目的/意义]社会化标注系统为用户检索提供便利的同时也面临一些困扰,标签推荐研究有助于解决资源检索中精确度与召回率之间的两难抉择.[过程/方法]借助网络科学的理论与方法,通过对标签网络的模块化聚类处理获得主题聚类,采用度数中心度对主题聚类内部标签进行排名,根据特定规则选取Top-N标签数量.[结果/结论]实验结果显示,研究中提出的模块化Top-N标签推荐方法,具有可逐层细化的精确度和良好的召回率.该方法的弹性机制可为不同的检索要求提供差异化服务.%[Purpose/Significance]Social tagging system is also facing some inconvenience while facilitating the re-trieval of users.Tag recommendation research can help solve the dilemma between the accuracy and the recall rate in re-source retrieval.[Process/Method]With the theory and method of network science, the topic clusters were obtained through the modularity clustering on tag network, the degree centrality was used to rank the tags in the topic clusters, the numbers of the Top-N tags were selected according to a specific rule.[Results/Conclusions]The experimental results showed that the modularity Top-N tag recommendation method had the accuracy of level-by-level refinement and the good recall rate.The flexible mechanism of this method could provide differentiated services for different retrieval requirements.