> Ralf Heidenreich wrote:
> > sa-learn coaches spamassassin.
On 23.01.09 10:45, Bowie Bailey wrote:
> Actually, sa-learn coaches the Bayes db if you want to be specific.
I prefer word "train" instead of "coach" :-)
> > Is it better, to coach spamassassin with mails, that are not examined
> > through spamassassin. Also original spam-mails.
> > If spamassassin examines mails, and writes a Spam-Status flag into the
> > header, can these mails used for sa-learn?
>
> It doesn't matter. Train with everything you have. sa-learn will
> automatically remove any SA headers. It will also automatically skip
> any messages that have been previously learned, so you don't have to
> worry about learning the same email twice.
However, most important is training on false-positives and false-negatives,
then on mail that hadn't been classified with high or low spam probablility,
especially mail that had too high (ham) or low (spam) score.
Yes, even training on BAYES_00 and BAYES_99 can give some advantages, but if
you don't have much time, fucos on those I described above
--
Matus UHLAR - fantomas, uh...@fantomas.sk ; http://www.fantomas.sk/
Warning: I wish NOT to receive e-mail advertising to this address.
Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu.
My mind is like a steel trap - rusty and illegal in 37 states.