Constructing Web User Profiles: A non-invasive Learning Approach

  • Authors:
  • Philip K. Chan

  • Affiliations:
  • -

  • Venue:
  • WEBKDD '99 Revised Papers from the International Workshop on Web Usage Analysis and User Profiling
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

Our web user profiles consist of Page Interest Estimators (PIE's) and Web Access Graphs (WAG's). We discuss a non-invasive approach to estimating the user's interest of a web page without directly asking the user. A time and space efficient method is proposed for locating multi-word phrases to enrich the common bag-of-words representation for text documents. PIE's are then learned to predict the user's interest on any web page. A WAG summarizes the web page access patterns of a user. We describe how a user profile can be utilized to analyze search results and recommend new and interesting pages. Our empirical results on PIE's are encouraging.