Study Of Personalized Human-Computer Interface Based On Agent

With the development and maturity of Internet and its relevant technology, Internet, especially the WWW increasingly becomes the daily platform where people released and obtained information. The way that people gathering information from the Internet has become more and more important nowadays and in the future.However, the online information likes the boundless ocean, an ordinary user uses the existing means to be very difficult to resist. This includes two kinds of situations: (1) Information that delivered to people voluntarily or downloaded is difficult to digest. This is named as "information overloading";(2) Users don't know how to express their demands for the web resources, or they don't know how to look for resource accurately and effectively. That is named as "resource wildering".This paper construct a system prototype based on intelligent agent and user model technology in the application field of "News Choosing System". This system finds out users' interest, demand and changes through observing, recording and analyzing the users' behavior, constructs personalized user model according to these interests and offer the information the users are interested in. This system helps users solve "information overloading" and "resource wildering" by doing the things above. This paper uses "intellectual Agent" as the tool of user's modeling. The autonomy, learning ability and communication ability of intelligent agent guarantee it can finish this task well.This system is comprised of user interface Agent, searching Agent, data analyzing and handling Agent and user model Agent. The user interface Agent is responsible for interacting with user, receiving user's command and reporting the result to user. Searching Agent is responsible for helping users look for the very content. Data analyzing and handling agent is responsible for news management, classifying ,learning reasoning and decision-making, etc.;User modeling Agent takes the responsibility of user modeling, maintaining and updating usermodel. This paper designed these Agents according to the structure of "mixed structure Agent", and achieved them by thread mechanism within Windows system. Because these four agents need to not only perform its own functions but also corporate to finish task, this paper designed an "Agent Message Processor" as the platform of agent communication. Four Agents don't communicate with each other directly, but Message Processor masters every agent's information and owns a "message reflection table" to know which agent can handle which kind of messages. Four Agents and Message Processor form a multi-agent system framework called "federation system".After the analyzing the classical cluster's algorithm, and improving algorithms, this paper proposes a method based on picture of the automatic cluster of user's interest.