cyrus and Spamassassin bayes rules

[Craig Ringer]

On Sun, 2004-04-25 at 20:58, Ghislain ADNET wrote:

>   To train the bayes filter of spamassassin you need to feed it the spam 
> and "ham" so it learn how to deal with it.
> It can learn email from an mbox unix file. Now you can see the problem 
> with cyrus as it does not use mbox format at all :)

sa-learn is quite happy reading Cyrus mail files, but to access them you
need to be user 'cyrus' or 'root'. This makes it impossible for normal
users, but it's quite reasonable for batch jobs.

I keep a public 'junk' folder that I run sa-learn on, and it's improved
our accuracy rates a lot. I use SpamAssassin via MimeDefang on a milter.
For "good" mail, I feed it a bunch of mailing lists, my personal
mailbox, and that of a few other staff who can be trusted to sort mail
instead of ignoring it.

>   So my question is, how can we do that ? do any of you experienced some 
> tools to convert a cyrus  mailbox to mbox format so it can be feeded to 
> spamassassin ?

Mailutil will do it, and is easy to script.

>  Or perhaps can i feed it directly from the actual 
> individual files in the mailbox ?

That's what I do.

Craig Ringer

---
Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html