This paper proposes methods by which user's preferences for WWW-based pages can be inferred from user's behaviors. Both explicit and implicit feedbacks of inference were used to infer the user's preferences. In the explicit feedback mode, a user evaluates the selected page as interest/not interest according to the relevancy of the page with the given query and sends an explicit feedback. In the implicit feedback mode, a user browses the pages by performing; for instances; bookmark, saving, printing, scrolling, enlarging, closing, reading, or jumping to another link, and the system infers from these operations how much the user was interested in the page. The users browse pages by using Kodama's simple browser in which there is an interaction agent that monitors the user behaviors and a learning agent that infers user's preferences from the interaction agent. The results show that the proposed techniques for learning and using user preferences in refining the given query and filtering the retrieved documents greatly enhance the value of retrieving more relevant information.
展开▼
