The Grumpy Editor's guide to bayesian spam filters
Posted Feb 23, 2006 1:54 UTC (Thu) by
mbcook (subscriber, #5517)
In reply to:
The Grumpy Editor's guide to bayesian spam filters by dlang
Parent article:
The Grumpy Editor's guide to bayesian spam filters
I have to agree. I've been using POPFile for years, both on Windows and OS X. It is a fantastic
little program. It doesn't detect "ham/spam", it puts things into buckets. Now I use Ham and
Spam as my buckets, but you can add more and it will learn where things go. So you could make
a bucket for kernel patches and it would learn when things go in there. That would probably
increase the accuracy since it doesn't have to lump kernel patches (which would be largely C
code) in with Ham (which would contain all sorts of stuff).
I recently tried the IMAP support, which was cool. I'm not an IMAP person so I went back to using
POP3 but it did have one really cool feature: classification by folders. Since it monitors what is in
each folder, it knows when you move things. Thus, when you move a message out of your Spam
folder into your Ham folder, it learns that it made a mistake (and vice versa). This is such a cool
idea. Because of the way it works you could set a little home server to run POPFile in IMAP mode
and then whenever and where ever you check your e-mail you can reclassify things and it will
learn from that.
It also supports SSL is you put in the have the needed perl modules. SSL support works great and
provides security that is nice to have when you often work on open networks where someone
could get your e-mail password.
Check it out.
(
Log in to post comments)
