TWIMC

Abstract

As the Internet and computers become commodities in our daily lives, electronic communications such as e-mails are rapidly substituting traditional postal mails. Although e-mails are more convenient, letters still have many positive features. The ability to handle "anonymous recipient" is one of them. This thesis proposes a software agent that performs the routing task as human beings for the anonymous recipient e-mails. The software agent named "TWIMC (To Whom It May Concern)" receives anonymous recipient e-mails, analyzes it, and then routes the e-mail to the most qualified person (i.e., e-mail account) inside the organization. We view each e-mail account as a category (or class) and everyday e-mail collections for each e-mail account provide an excellent source of training data. We also propose a novel algorithm that generates a robust and efficient classifier, named Set-oriented GA Classifier System (SCS). It uses set representation for the classifier and inverted file for fast fitness evaluation. We compare SCS with Support Vector Machine (SVM), which recently draws interest of text categorization researchers. The result shows that the SCS outperforms SVM under noisy environment and TWIMC can be used in the real situation.