Building Enhanced Link Context by Logical Sitemap

摘要

Link contexts have been applied to enrich document representation for a variety of information retrieval tasks. However, the valuable site-specific hierarchical information has not yet been exploited to enrich link contexts. In this paper, we propose to enhance link contexts by mining the underlying information organization architecture of a Web site, which is termed as logical sitemap to differ from sites supplied sitemap pages. We reconstruct a logical sitemap for a Web site by mining existing navigation elements such as menus, breadcrumbs, sitemap etc. It then enriches contexts of a link by aggregating contexts according to the hierarchical relationship in the mined logical sitemap. The experimental results show that our proposed approach can reliably construct a logical sitemap for a general site and the enriched link contexts derived from the logical sitemap can improve site-specific known item search performance noticeably.