IEMS - an approach that combines handcrafted rules with learnt instance based rules

Sorting email messages into folders is an important task for a vast number of computer users. The i-ems (Intelligent-Electronic Mail Sorter) project is focused on improving the way users interact with email managers. In this article we report on two aspects of the i-ems project. Firstly, we discuss how combining learnt and hand crafted rules can improve the overall performance of the system. Secondly, we present a composite rule learner that classifies mail by combining an instance based approach with an approach that constructs a general explicit description. We show that this approach improves classification performance. Furthermore, this combined approach produces understandable and concise classification rules that users can scrutinize allowing them to maintain a sense of control.