一种基于用户配置文件的个性化检索方法  

Personalized retrieval method based on user profile

作　　者：胡妙[1] 戴牡红[1] 陈浩[1] 

机构地区：[1]湖南大学信息科学与工程学院,长沙410082

出　　处：《计算机应用研究》2016年第2期417-420,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61472132)

摘　　要：为了进一步改进个性化搜索方法,通过对现有个性化搜索方法的研究,提出了一种新的搜索方法。该方法从用户兴趣相关性出发,将用户配置文件与传统个性化搜索相结合。在传统TF-IDF方法的基础上,提出了一种综合考虑标签总引用次数和配置文件中标签总数的新方法,用于获取用户配置文件与资源配置文件中的标签权重;设计了基于余弦相似性计算并综合匹配的标签个数的资源相关性计算方法。通过Movie Lens数据集实验,验证了提出方法的有效性。In order to further improve personalized search methods, by researching existing personalized search methods, this paper proposed a new search method. The method started from the user＇ s interest correlation and combined the user profile with the traditional personalized search. Based on the traditional TF-IDF method, it put forward a new comprehensive method which took the total number of tags in the configuration flies and the total citations of a tag into consideration to get the weight of the tags in the user profile and resource profile. Then it designed a correlation calculation method of resources, which based on the cosine similarity calculation and took the number of matching tags into consideration. Experimental results with Mov- ieLens data set verify the validity of the proposed method.

关 键 词：个性化搜索 标签 配置文件 相关性匹配 余弦相似性 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]