Dynamic modeling and learning user profile in personalized news agent

Abstract

Finding relevant information effectively on the Internet is a challenging task. Although the information is widely available, exploring Web sites and finding information relevant to a user's interest can be a time-consuming and tedious task. As a result, many software agents have been employed to perform autonomous information gathering and altering on behalf of the user. One of the critical issues in such an agent is the capability of the agent to model its users and adapt itself over time to changing user interests. In this thesis, a novel scheme is proposed to learn user profile. The proposed scheme is designed to handle multiple domains of long-term and short-term users' interests simultaneously, which are learned through positive and negative user feedback. A 3-descriptor interest category representation approach is developed to achieve this objective. Using such a representation, the learning algorithm is derived by imitating human personal assistants doing the same task. Based on experimental evaluation, the scheme performs very well and adapts quickly to significant changes in user interest.