Syskill and Webert Web Page Ratings

Task Type

classification

Sources

Donor

Michael Pazzani
Department of Information and Computer Science,
University of California, Irvine
Irvine, CA 92697-3425
pazzani@ics.uci.edu

Date Donated: October 20, 1998

Problem Description

The problem is to predict user ratings for web pages (within a subject category). The HTML source of a web page is given. Users looked at each web page and inidated on a 3 point scale (hot medium cold) 50-100 pages per domain. However, this is realistic because we want to learn user profiles from as few examples as possible so that users have an incentitive to rate pages.

The accuracy of predicting ratings is reported in early publications. Later publications used the precision at top N or the F-measure.

Results

The initially study compared traditional meachine learning methods with IR methods. A variety of learning algorithms worked acceptably, including naive bayes, nearest neighbor, and Rocchio's method

Pazzani M., Billsus, D. (1997). Learning and Revising User Profiles: The identification of interesting web sites. Machine Learning 27, 313-331

Pazzani, M., Muramatsu J., Billsus, D. (1996). Syskill & Webert: Identifying interesting web sites. Proceedings of the National Conference on Artificial Intelligence, Portland, OR. PDF

The UCI KDD Archive
Information and Computer Science
University of California, Irvine
Irvine, CA 92697-3425

Last modified: July 12, 1999