Mike Woods wrote:
> Bowie Bailey wrote:
> >
> > You don't have to figure out what has and has not already been
> > learned. Just filter everything into directories for ham and spam
> > learning and feed the entire directory to sa-learn. Once you have
> > learned the messages, you can either delete them, or move them to
> > an archive directory. Using this method, nothing is ever learned
> > twice. Also, (if you do your sorting properly) nothing is ever
> > learned incorrectly.
>
> That's basicly what I have in mind!
You can put this:
bayes_auto_learn 0
into your local.cf for global usage, or into a user's user_prefs as a
single user setting. This will completely turn off the auto learning.
Then you can sort the messages and learn them manually without having
to worry about what auto-learn is doing.
(This is for SA 3.1x, earlier versions may have a slightly different
config setting)
--
Bowie
