Maintenance Best Practices
Adam Mercer
r.a.mercer at blueyonder.co.uk
Thu Jun 5 21:32:29 CEST 2003
On Thu, Jun 05, 2003 at 03:14:20PM -0400, David Relson wrote:
> >I initially train my database with, at the moment, around 4000 known
> >spams and 4000 known hams. Then I train only on mistakes. So far this
> >seems to be working quite well with only a few false negatives - no
> >false positives yet.
>
> You're getting excellent results. Glad to hear that the default values are
> doing so well for you.
>
> Bogofilter _may_ be able to do even better for you. The new stable
> release, bogofilter-0.13.6.1, has revised directions on how to find, given
> _your_ email, the optimum combination of the spam_cutoff, min_dev, and robs
> parameters. If you have your messages in readily available .mbx files, you
> may want to experiment with the scripts in bogofilter/tuning.
I based this training on Gregs tuning howto, I'll take a look at
0.13.6.1
Cheers
Adam
More information about the Bogofilter
mailing list