A Multi-Agent System for Information Acquisition from the Web and its Dissemination

Sufang Hou

Abstract

This talk will discuss open problems in the design of a multi-agent system for information acquisition and dissemination. The talk will give an overview of the agent concepts, give a classification of agents, and describe the existing architecture of the multi-agent system. This multi-agent system integrates content-based and opinion-based information filtering capabilities to improve the quality and relevance of items delivered to the user. In the main body of the talk, I will present the expected contributions of my thesis:

1. A full implementation of the social trust model and user-profile update mechanism used in opinion-based filtering.
2. Analyze sources of overhead in the current system and improve its performance
3. Deal with privacy and security aspects of the system. For example, it should be possible for users to control their profiles and decide which information is conveyed to another party, and sensitive user information (e.g., their ratings of other users should be suitably protected).