Text Categorization Using a Personalized, Adaptive, and Cooperative MultiAgent System

In this paper, a multiagent system for supporting users in retrieving information from heterogeneous data sources, and classifying them according to users' personal preferences, is presented. The system is built upon PACMAS, a generic architecture that supports the implementation of Personalized, Adaptive, and Cooperative MultiAgent Systems. Preliminary tests have been conducted to evaluate the effectiveness of the system in retrieving and classifying newspaper articles. Results show an avarage accuracy of about 80%. I. I NTRODUCTION The information available on the WWW is continuously growing from different points of view: information sources are increasing, topics discussed are becoming more and more heterogeneous, and stored data has reached a considerable size. It has become a difficult task for Internet users to select contents according to their personal interests, esp ecially if contents are continuously updated (e.g., news, newspaper articles, reuters, rss feeds, blogs, etc.). Unfortunately , tradi- tional filtering techniques based on keyword search are ofte n inadequate to express what the user is really searching for. Furthermore, users often need to refine by hand the achieved results.