Bernard Merialdo, Joachim Soderberg, Diana Marras

Part of this project was done in cooperation with TELIS, Sophia-Antipolis, France, and the DCN Laboratory, Toulon, France.

Project description

The objective of this project is to design and evaluate techniques for the automatic filtering of email. In particular we have investigated the use of probabilistic models and dimensionality reduction techniques for this task. Some of this work was integrated in a project for an advanced email software, which included both numerical and symbolic approaches to the the filtering task. We specifically studied training procedures where the system learns automatically from informations provided by users.