IEMS - The Intelligent Email Sorter

Classification of email is an important everyday task for a large and growing number of users. This paper describes the machine learning approaches underlying the i-ems (Intelligent-Electronic Mail Sorter) system. The system has two distinctive aspects: if offers a view of the inbox based on predicted classifications of messages; and it provides the user with details of the accuracy of the predictions and the processes underlying them. We introduce a composite rule learner which classifies mail by combining an instance based approach with an approach which constructs a general explicit description. This was devised in order to achieve understandable, concise and effective classification rules. We report results of this learner as well as several others for data from for five users who apply very different ways of classifying their email. We discuss the implications of our results, in terms of the performance of the learning approaches, their sensitivity to concept drift and the ease w...

• Type: Conference
• Year: 2002
• Where: ICML
• Authors: Elisabeth Crawford, Judy Kay, Eric McCreath
• Keywords: Composite Rule Learner, Effective Classification Rules, ICML 2002, Machine Learning, Machine Learning Approaches