Data Discovery on the Information Highway (includes ProFusion and ProFilter)
Project Award Date: 0000-00-00
The Information Highway has become a reality. Increased access to the Internet by the public at large, combined with the development of easy-to-use graphical browsing interfaces-for example, Mosaic and Netscape-has led to an explosion of added information. The World Wide Web (WWW) in particular is being used to present an exponentially growing amount and range of information which people can browse through. Unfortunately, "too much information" can be the same thing as "not enough information": if the information you seek is buried under an avalanche, is it really there? This research has developed a pilot system which can automatically browse the on-line information, filter it, and display it, bringing the information to the user rather than requiring the user to seek it out.
In the first phase of this project, we developed a meta search engine, ProFusion. The goal of this project was to develop an intelligent, adaptive Web search tool. ProFusion, was awarded the PC Magazine Editor's Choice Award in the "Advanced Meta-Search" category for 1997.
ProFusion analyzes incoming queries, categorizes them, and automatically picks the best search engines for the query, based on a priori knowledge (confidence factors) which represents the suitability of each search engine for each category. It uses these confidence factors to merge the search results into a re-weight list of the returned documents, removes duplicates and, optionally, broken links, and presents the final rank-ordered list to the user.
The main goals of this phase of the research were to (1) provide ProFusion with a multi-agent architecture which is easier to extend, maintain, and distribute; and (2) to include automatic adaptation algoritms to replace the hard-coded a priori knowledge. Our multi-agent architecture demonstrates various desirable agent characteristics, including task-oriented modules, task-specific solutions, de-emphasized representations, decentralized control structure, and learning and development.
The second phase of our information filtering project, ProFilter, uses a database to store queries (and their results) for individual users. Over time, the queries are automatically re-run and the new results added to the user's collection. User feedback can be used to eliminate irrelevant results from those displayed by the user. the irrelevant data, however, remain in the database to prevent their being presented to the user as the result of a leter search. We are conducting research on how to incorporate intelligent pre-filtering of the search results based on user feedback into the operational system.
Primary Sponsor(s): NSF and KU Faculty Development Fund