Next: Results Up: Specializing to User Previous: Scenario 1

Experiment 1

The only agent in the system is a ``China'' agent, with a population of one profile. Effectively, the GA is not used at all and only the feedback learning system is tested. The initial profile is nearly empty and is shown on the left hand side of Table . The system is iteratively used for a number of user sessions. The number of articles shown to the user is fixed and the articles are presented in a prioritized order. Each session involves reading the articles retrieved by the profile and providing positive feedback for articles about China.

Note that the location fields are not used in this experiment nor are individual keywords selected while providing feedback. The only feedback provided is for the document as a whole. This constraint is imposed so that the result can hold for most situations. In most filtering environments, at least the text of the article is always available. Therefore, no pre-indexed fields are used to avoid loss of generality.


MIT Media Lab - Autonomous Agents Group - agentmaster@media.mit.edu