Abstract: How to identify and evaluate information credibility ranking has become an increasing important problem. To address the issue， an effective credibility evaluation method called C-Rank to compute trust values of records in Deep Web databases is proposed， which constructs an S-R credibility graph for each record. The graph contains 2 types of vertices and 3 types of edges. Firstly， each vertex’s local trust value is computed by using out-degrees based on the idea of trust propagation. Then， the weight of record vertex is computed by using its in-degree and adjacent site vertices’ local trust values. Lastly， the global trust value of the S-R credibility graph is computed， which denotes the record’s credibility in the whole Web. Experiment results show C-Rank can evaluate credibility rankings of records appropriately and discriminate false information effectively. This method is generally applicable to all domains of Deep Web.

Key words: Deep Web, Web information credibility, S-R credibility graph, trust propagation

摘要： 针对Web信息可信度问题，提出了一种为Deep Web数据记录计算可信度的有效方法C-Rank。该方法为每一条记录构造一个S-R可信度网络，包含两种类型顶点及三种类型边。首先基于可信度传播的思想，利用顶点出度为每一个顶点计算其局部可信度值；再利用Record顶点入度及相邻Site顶点的可信度值，为该Record顶点计算权值；继而求得整个S-R网络的全局可信度值。实验证明，C-Rank方法能够合理而有效地评价数据记录的可信度，从而达到甄别虚假信息，为用户推荐可信数据记录的目的。该方法普遍适用于Deep Web的各个领域。

关键词: 深层网络, Web信息可信度, S-R可信度网络, 可信度传播