A multi-agent system for information retrieval in the Web using the fuzzy bayesian rule

A multi-agent system based on a fuzzy bayesian approach aimed to help internet users to filter Web pages is discussed. In this approach, one page can be classified as having the continuous quality of being interesting, which means that a certain grade of membership can be associated with each page relative to a category of selection. Filtering is based on the evidences of the content of the page title, abstract or complete document. The multi-agent system will search for and retrieve from, on several information and data sources spreaded on the Web, those documents that might be interesting to the user; such a task lead to unefficient and uneffective results using conventional searching methods. An example comparing crisp and fuzzy classifiers implemented as a part of multi-agent system to support information filtering and retrieval in the Web is discussed illustrating the proposed approach.