| This thesis identifies the information overload problem as a result of the increasingly vast amounts of heterogeneous content available on the Web. As an attempt to meet the challenge, a multi-agent system is proposed to improve the efficiency and effectiveness of relevant knowledge acquisition and dissemination from various Web information sources. We analyze and design the agent system using pattern-driven approach. The system consists of a group of agents that work collaboratively to achieve goals that are beyond the scope of an individual agent. The multi-agent system locates and retrieves relevant information from various Web data sources according to users' profiles, performs opinion-based filtering on the information to improve its quality, integrates and presents the information to users.; A prototype of the multi-agent system is implemented to demonstrate the capabilities of the proposed system. AgentMom is adopted as the framework to develop the physical agent system. |
