On extracting link information of relationship instances from a web site

Abstract

Web pages from a web site can often be associated with concepts in an ontology, and pairs of web pages can also be associated with relationships between concepts. With such associations, web pages can be searched, browsed or even reorganized based on their concept and relationship labels. In this paper, we investigate the problem of extracting link information of relationship instances from a web site. We define the notion of link chain and formulate the link chain extraction problem. An extraction method based on sequential covering has been proposed to solve the problem. This paper presents the proposed method and the experiments to evaluate its performance. We have applied the method to extract link chain information from the Yahoo! Movie Web Site with very promising results. © Springer-Verlag Berlin Heidelberg 2003.