基于标准标签的用户兴趣模型研究  

Research on User Interest Model Based on Standard Tag

作　　者：杨晶[1] 成卫青[1] 郭常忠[2] 

机构地区：[1]南京邮电大学计算机学院,江苏南京210003 [2]烟台大学数学与信息科学学院,山东烟台264005

出　　处：《计算机技术与发展》2013年第10期208-211,215,共5页Computer Technology and Development

基　　金：国家自然科学基金资助项目(61170322;71171117);软件开发环境国家重点实验室开放课题(SKLSDE-2011KF-0X);江苏省自然科学基金资助项目(BK2010524)

摘　　要：信息大爆炸的网络时代,个性化推荐是解决信息"超负载"的有效办法。用户兴趣模型是个性化推荐的核心,关系着整个推荐系统的推荐质量。标签一直被用于资源分类,在个性化推荐方面却很少使用。文中采取向量空间模型的建模方法,利用个性化标签描述用户兴趣,并提出一套简洁有效的标签标准化方法—基于属性共现率的标签标准化以及基于聚类的标签标准化方法对用户的自定义标签进行标准化。该模型能有效降低用户兴趣模型的向量维数,避免分析标签语义的复杂过程,且能够从用户的角度贴切地表达用户兴趣。实验结果表明该模型有助于提高个性化推荐的服务质量。Faced to the Internet age of information explosion,the personalized recommendation is an effective way to solve the“informa-tion overload”. User interest model as the core of personalized recommendation determines the quality of the recommendation system. Tags have been used for the classification of resources;however,they are seldom used in personalized recommendation. In this paper,vec-tor space model is used in modeling,where personalized tags are used to describe user interests. A set of simple and effective methods are proposed to standardize user＇s custom tags,including a standardization method based on attribute co-occurrence frequency and a stand-ardization method based on clustering. Thus,the vector dimension of the user interest model can be reduced effectively,avoiding complex tag semantic analysis,as well as being able to aptly express user＇s interests from their point of view. The experimental results show that the proposed user interest model can help to improve the quality of personalized recommendation.

关 键 词：个性化推荐 用户兴趣模型 向量空间模型 标准标签 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]