Evaluation of Web Retrieval Methods Using Anchor Text

In this paper, we evaluate two types of anchor texts: a page anchor and a site anchor. Since the anchor text tends to summarize information referred ahead, it can be expected that the terms appearing there have important meaning in information retrieval. We introduce a retrieval method to give high priority to the terms in the anchor text. In the experiment, we compared the proposed method with the base line which indexed only page documents. The result indicated that both methods had almost the same accuracy, and that there were many queries in which the accuracy much differed between two methods. It can be expected that the improvement on the queries in which the proposed method was inferior to base line will be achieved by deleting overlapped anchor texts toward the same page.