On Wed, 2 Jun 2004, Duncan Hill wrote:
> On Tuesday 01 June 2004 18:31, Spam Admin might have typed:
>
> > > What score does Bayes assign?
> > > Have you told Bayes that these mails are spam?
> >
> > It's doubtful I'm using Bayes, as this is simply a "filter and forward"
> > server to an internal GroupWise system.
>
> Worth poking with a global bayes db. It's not as good as a per-user, but it
> still works pretty well.
>
I will second that wholeheartedly. I'm running SA on a front-end
relay with global bayes and auto_learning and it's mostly right.
(when it's not bang on, it's usually just dithering at 50% until
it see enough examples of the particular spam/ham to train).
auto_learning works best if you also use some other dynamic component
in your SA to help it identify new spam (DNSBLs, dcc, razor, surbl, etc).
Lacking those, you might need to spend some time manually training your
Bayes.
--
Dave Funk University of Iowa
<dbfunk (at) engineering.uiowa.edu> College of Engineering
319/335-5751 FAX: 319/384-0549 1256 Seamans Center
Sys_admin/Postmaster/cell_admin Iowa City, IA 52242-1527
#include <std_disclaimer.h>
Better is not better, 'standard' is better. B{
