Structured Information Retrieval for Web Documents

To overcome the limitations of conventional Web search engines in retrieving Web documents relevant to users' queries, one has to exploit semantic structures embedded in Web documents. We propose a Web Information Retrieval (WebIR) model for Web documents containing semantic elements which are text segments enclosed by special tags. These special tags, known as semantic tags, can either be independently created for individual Web documents, or be standardized for a collection of Web documents sharing common types of semantic elements. The WebIR model supports queries on both intra-document semantic elements and inter-document links, and returns directed graphs as query results. Each directed graph represents a cluster of connected Web documents satisfying the given query. The collection of directed graphs in a WebIR query result is further ranked. In this paper, we describe the WebIR model and an ongoing implementation eeort to realize the model.