DaimlerChrysler maps US highways by tracking GPSs in cars. This project proposes an analogous approach for the Web by discovering routes through a website from click-stream data in a Web server log. It seeks to add intelligence to a standard website by combining Web Usage Mining of click-stream data for incoming users, with Reinforcement Learning to identify optimal routes within the website. Web mining identifies users' needs and information sources from both plentiful usage click-stream data and knowledge of the website's content and structure. Mining the usage data identifies frequently followed paths through the website. The intelligent website will react autonomously to changes in access patterns as well as evolving website content and organisation. Usage knowledge will improve web browsing through a virtual website guide or by intelligent manipulation of the website.s hyperlinks. This project provides one of the three "smart" targets of RGU's SmartWeb Technologies Centre — a "smart" repository, rather than a "smart" interface or Web application.    [More]

Funding

• RGU RDI

Research Team

• Susan Craw (PI)
• Ganesan Bathumalai
• Frank Herrmann

Related Projects

• Intelligent Management of Email
• SmartWeb — a centre of excellence for intelligent Web technologies
• AmbieSense — ambient, personalised, and context-sensitive information systems for mobile users
• ProfileSkim — an intelligent document browser
• Active Groups — knowledge sharing tools for email
• Facilitating Information Retrieval By User-Adaptive Learning
• WebCluster — mediated access to the web

Publications

Craw, S. (1999). Reinforcement learning for information seeking. IEE Seminar on Searching for Information: Artificial Intelligence and Information Retrieval Approaches