Simplifying Web Traversals by Recognizing Behavior Patterns

Web sites must often service a wide variety of clients. Thus, it is inevitable that a web site will allow some visitors to find their information quickly while other visitors have to follow many links to get to the information that they need. Worse, as web sites evolve, they may get worse over time so that all visitors have to follow many links to find the information that they need. This paper describes an extensible system that analyzes web logs to find and exploit opportunities for improving the navigation of a web site. The system is extensible in that the inefficiencies that it finds and eliminates are not predetermined; to search for a new kind of inefficiency, web site admininstrators can provide a pattern (in a language designed specifically for this) that finds and eliminates the new inefficiency.