基于用户兴趣模型两段式排序的隐私保护方法

针对现今个性化搜索中用户兴趣模型隐私的保护需要,提出一种基于服务器和客户端协作的用户兴趣模型两段式排序方法.利用基于上下层次结构的树状用户兴趣模型,对其分段在服务器与客户端对搜索结果进行排序,不仅提高了个性化搜索服务质量,而且通过用户可控的开放隐私程度调节,有效地达到了隐私保护的目的;此外,该模型采用动态目录结构,实现了用户兴趣数据的反馈更新,从而使得用户兴趣模型的信息更加精确.实验结果表明:该模型的个性化搜索质量优于Google原始排序结果,这种差距随着用户隐私开放控制程度的提升逐渐变小;在服务器两段排序对中间排序结果的裁剪过程中,只要恰当地选择数据的范围,不仅不会影响个性化搜索服务质量,而且能提高系统运行效率.

Privacy protection method based on user profile of two-step sorting

A two step sorting method based on the corporation between server and client was proposed to meet the requirement of protecting the privacy in user profile in  personalized web search. By using the user profile based on concept tree structure, the method sorted the results on the server and client respectively, which both improved the quality of personalized web search and protected the privacy of the user by controlling the open privacy parameter. Whats more, this user profile of dynamic concept structure could be updated in time to make the data more accurate. The experimental results show that the search quality is much better than  Googles and the difference decreases when users enhancing their privacy. During the process when server pruning the temporary results, if  the scope is selected appropriately, the search quality will not be influenced and the system performance can even be improved．