Using web helper agent profiles in query generation

  • Authors:

  • Gabriel L. Somlo;Adele E. Howe

  • Affiliations:

  • Colorado State University, Fort Collins, CO;Colorado State University, Fort Collins, CO

  • Venue:
  • AAMAS '03 Proceedings of the second international joint conference on Autonomous agents and multiagent systems
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Personalized information agents can help overcome some of the limitations of communal Web information sources such as portals and search engines. Two important components of these agents are: user profiles and information filtering or gathering services. Ideally, these components can be separated so that a single user profile can be leveraged for a variety of information services. Toward that end, we are building an information agent called SurfAgent; in previous studies, we have developed and tested methods for automatically learning a user profile [20]. In this paper, we evaluate alternative methods for recommending new documents to a user by generating queries from the user profile and submitting them to a popular search engine. Our study focuses on three questions: How do different algorithms for query generation perform relative to each other? Is positive relevance feedback adequate to support the task? Can a user profile be learned independent of the service? We found that three algorithms appear to excel and that using only positive feedback does degrade the results somewhat. We conclude with the results of a pilot user study for assessing interaction of the profile and the query generation mechanisms.