POPFile classifies email into categories you define. It can sort into spam and not spam or into any number of categories you like (e.g. work, personal, important project, hobby, etc.).
The classification is done using a naïve Bayes algorithm. In other words, POPFile uses statistics to track which words are likely to appear in which messages. This means that POPFile will adapt to the kind of mail you receive and needs to be trained. Out of the box, it doesn't know anything about spam or how messages from your mother differ from those your friends send you. However, if you train it, it will soon learn how to tell these different kinds of messages apart.
POPFile was created by John Graham-Cumming and is maintained as open source by a team of developers and users that file bug reports, post patches, and suggest enhancements.