社会化标注系统中基于标签和项目的兴趣建模比较研究  

Comparison Study on Tag-Based and Item-Based User Profiles Modeling in Social Tagging Systems

作　　者：胡潜[1] 林鑫[2] 

机构地区：[1]华中师范大学,武汉430079 [2]武汉大学信息管理学院,武汉430072

出　　处：《情报学报》2015年第12期1296-1303,共8页Journal of the China Society for Scientific and Technical Information

基　　金：国家自然科学基金项目数字图书馆社区的知识聚合与服务研究(项目编号:71273197)的研究成果之一

摘　　要：近年来，基于标签构建用户兴趣模型受到了重点关注，然而也有研究对其有效性提出了质疑。为验证标签是否适合作为兴趣建模的基础数据、效果是否更加理想，本文以电影社会化标注系统为例，采用空间向量的方法分别基于标签和项目进行了兴趣建模，并以用户召回率、推荐项目召回率和准确率为评价指标进行了效果比较分析。结果显示，基于标签建模策略效果显著差于基于项目建模策略；但是在用户标注影片较多的情况下，基于标签建模策略也取得较好效果。研究的局限性主要表现在两个方面：一是分析对象选用的是影片，其结果未必适用于音乐、图书、网页等类型的资源；二是研究仅针对影片类型兴趣进行了建模，没有建立综合模型，因而其结果可能未全面反映基于标签和项目建模策略的效果。In past years, a number of studies focus on tag-based user profiles modeling, but its effectiveness is questioned by some studies. To verify whether social tags are suitable for user profiles modeling, and whether tag-based user profiles＇ effectiveness is better, this paper takes movies as sample, uses space vector model to construct tag-based and item-based user profiles, and compares their effectiveness by the following three evaluation indexes： user recall, item recall and precision. Based on the results, it is can be concluded that the effectiveness of item-based user profiles is apparently better than tag-based user profiles; but tag-based user profiles for users who tagged enough movies performs well. Limitations of this study are as follows： this study only takes movies as sample, and they may not reprehensive for other types of resources, e.g. music, books and webpages; user profiles only cover users＇ interests on movie genres, and the results may not comprehensively reflect the effectiveness of tag-based and item-based user profiles.

关 键 词：社会化标签 用户兴趣 建模个性化 

分 类 号：G252[文化科学—图书馆学]