Personal rules generation for intelligent e-mail service system

Abstract

As E-mail becomes more popular over network, the problem of receiving a lot of undesired E-mails arise. All of the previous systems, which have been developed to solve this problem, may be divided into two different categories, client-side, and server-side, from the view of location of filter. The former case always causes waste of network traffic and the latter case results that ail users may only share a small predefined set of filtering rules. In this paper, we propose a new architecture of E-mail service system, called Intelligent E-mail Service System (IESS), integrating client-side and server-side to help users manage their E-mails. A data mining approach is used in client to find managing rules from user's behavior of reading E-mail automatically. Then these rules would be stored in rule base for predicating E-mail at server. The prediction of the E-mail is made before transferring it to client; therefore, the undesired E-mails can be filtered at server to save network traffic. In addition, the feedback mechanism is provided to refine the prediction accuracy based on each user's decisions for reading E-mails. Experimental results show that IESS can infer E-mail reading preference of individual users correctly up to the accuracy about 80%. We may conclude that IESS can help users manage their E-mails effectively.