On Tuesday 29 June 2004 10:35 am, Andy Norris wrote:
> My setup is site-wide, and I have trained Bayes with a couple thousand
> spam.

You also need to make sure it's trained with at least 200 non-spam messages.  
Otherwise, it has nothing to compare the spam signs against.

Take some sample legit mail and run "sa-learn --ham" on it, or enable 
auto-learn for a while, and Bayes should start running.

-- 
Kelson Vibber
SpeedGate Communications, <www.speed.net>

Reply via email to