基于社会化标注的用户兴趣发现及个性化推荐研究

基于标签的个性化推荐应用越来越普遍，但是标签带有的语义模糊、时序动态性等问题影响着个性化推荐质量，现有研究仅从数量和结构上考虑用户与标签的关系。基于社会化标注系统的个性化推荐首先对融合社会关系的标签进行潜在语义主题挖掘，然后构建多层、多维度用户兴趣模型，提出模型更新策略，最后实现个性化推荐。采集CiteUlike站点数据进行实验分析，结果表明改进算法比传统算法更准确表达用户兴趣偏好，有效提高了个性化推荐准确率。

Research on User Interest Mining and Personalized Recommendation Based on Socialized Annotation

Tag-based personalized recommendation is becoming more and more popular,but the semantic ambiguity and temporal dynamics of the labels affect the quality of personalized recommendations,and the existing researches only consider the relationship between users and labels in terms of quantity and structure.Personalized recommendation based on social tagging system firstly mined potential semantic topics of the tags that fused social relationships,then built multi-level,multi-dimensional user interest models,and proposed a model update strategy,finally achieved personalized recommendations.Collecting CiteUlike site data for experimental analysis,and the results showed that the improved algorithm expressed user interest preferences more accurately than traditional algorithms,and effectively improved the accuracy of personalized recommendation.