Chapter 5 INTELLIGENT AGENTS FOR DOCUMENT CATEGORIZATION AND ADAPTIVE FILTERING USING A NEURAL NETWORK APPROACH AND FUZZY LOGIC Frank Teuteberg Business Informatics, European University Viadrina, Frankfurt (Oder), Germany Abstract: This chapter presents a multi-agent system for document categorization and adaptive filtering. A neural network approach is proposed to automate the process of agent-based categorization and adaptive filtering of electronic in- formation sources from the WWW. Results from training and testing those networks are presented and discussed. Fuzzy logic is used to handle impre- ciseness in agent communication and collaboration. Key words: Fuzzy Logic, Information Filtering, Intelligent Agents, Multi-Agent Systems, Neural Networks, XML 1. INTRODUCTION The World Wide Web (WWW) is semi-structured and dynamically changing. The number of WWW sites is huge and still growing fast. There- fore, it can be a time consuming task to find the appropriate information sources. Keyword-based search engines are often not of much help whenever qualitative criteria are important to describe the user’s information needs. For this reason, it is desirable that information agents support the extraction, filtering and categorization of information in the WWW. The problem our research is motivated by is to find e-commerce applica- tions which are “interesting” from a business point of view. “Interesting” means, for example, that a WWW site provides support for business transac- tions. One way to find “interesting” e-commerce applications is using a