中文摘要：

用户特征的描述方式是实现个性化搜索算法的核心因素。针对传统的基于关键词向量空间模型的用户特征描述过于简单,不能全面描述用户兴趣的缺陷,将folksonomy的结构与本体概念的清晰语义相结合,提出一种多层用户特征描述方式。从用户兴趣主题、用户间关联两个不同角度,从用户生成的标签、标记的文档及主题等不同层次建立用户特征描述模型,并将其应用于个性化搜索过程的方式进行分析。同时对个性化搜索的结果评价方式、资源类型对用户特征及搜索结果的影响进行了讨论。在Delicious和Flickr两种不同类型数据集上的实验表明,所提出用户特征模型能够有效提高个性化搜索结果的性能。

英文摘要：

User profile description is the core component of personalized search algorithm.The traditional keywords based vector space model is too simple to describe the user profile completely.By combining the structure of folksonomy and the explicit semantic of ontology concepts, a multi-layer user profile description model is proposed.Both user interesting and user relation are taken into considered in the model, and user profile can be described in different abstract layer such as tag, documents annotated and topics.Evaluation method of personalized search and the effect of resource type on user profile are also discussed.Experiments on Delicious dataset and Flickr dataset show that the user profile model proposed can improve personalized search performance effectively.