In addressing the growing problem of SPAM e-mail on the Internet, we examine methods for the automated construction of filters to eliminate such unwanted messages from user's mailbox. We take notice that e-mail filtering is a text classification task....
In addressing the growing problem of SPAM e-mail on the Internet, we examine methods for the automated construction of filters to eliminate such unwanted messages from user's mailbox. We take notice that e-mail filtering is a text classification task. The main problems in text classification are lack of labeled data, as well as the cost of labeling the unlabeled data. We address these problems by adapting co-training. We experiment with adaptive co-training on the e-mail domain. Our results show that the performance of adaptive co-training depends on the learning algorithm and features it uses. In particular, adaptive co-training outperforms co-training and other algorithms on email classification.