Automatically Constructing Descriptive Site Maps

Rapid increase in the number of pages on web sites, and widespread use of search engine optimization techniques, lead to web sites becoming difficult to navigate. Traditional site maps do not provide enough information about the site, and are often outdated. In this paper, we propose a machine learning based algorithm, which, combined with natural language processing, automatically constructs high quality descriptive site maps. In contrast to the previous work, our approach does not rely on heuristic rules to build site maps, and does not require specifying the number of items in a site map in advance. It also generates concise, but descriptive summaries for every site map item. Preliminary experiments with a set of educational web sites show that our method can construct site maps of high quality. An important application of our method is a new paradigm for accessing information on the Web, which integrates searching and browsing.

• Type: Conference
• Year: 2006
• Where: APWEB
• Authors: Pavel Dmitriev, Carl Lagoze
• Keywords: APWEB 2006, Internet Technology, Site Maps, Traditional Site Maps, Web Sites