[Jscheme-user] A plan for Spam

[Ken A. <kan...@bb...>]

I implemented Paul Graham's spam filter:
http://www.paulgraham.com/spam.html.

I'm impressed at how simple (< 160 lines) and effective it is.

For training i used 290 spams and the first 500 messages in my inbox (which includes spam, though only 2 in the first 500.

I then ran it on the 3200 messages in my inbox and it found 313 spams.
There was one border line false positive - a Java Development bulk mailing.
Not bad for 2 - 3 hours work!

The code is Eudora specific, but you should be able to tailor it to your mail reader easily.  It can classify about 10 emails/sec on my machine, which is plenty fast enough for playing around.

Thanks, Paul!

k