Analysis of Topics and Relevant Documents for Navigational Retrieval on the Web

An investigation into relationships among search methods, topic types, and styles of relevant Web pages was conducted using the "NTCIR-4 WEB Test Collection for Navigational Retrieval", which was built for evaluating the retrieval effectiveness of Web search systems aiming to find representative web pages of a given object. First, "Known Item Search," the task of the test collection, is defined and its outline is given. It consists of 100 gigabytes of Web document data NW100G-01, 300 topics of various types, and corresponding relevance judgments. Next, summaries of the systems and their evaluations used in this investigation are shown. Some systems used anchor text information or link information, and others used only the content information in various ways. Finally, characteristics of topics and their relevant documents and some interesting observations are discussed in relation to search methods and their effectiveness.