On Fri, 2009-07-31 at 07:53 -0500, Dennis B. Hopp wrote:
> I know that the bayes success rate comes down to training, but like
> every other administrator I can't possible check every message for
> accuracy and I was hoping to make the auto learn a little better. I
> thought maybe I just didn't have enough rules (both negative and
> positive scoring) to trigger the auto learn often enough.
As I wrote before, you particularly need to train spam with a low-ish
Bayes confidence, regardless of the overall SA score or the number of
rules hit. This does need some supervision.
One way to help this would be to set up some (server-side) rules to
deliver all spam triggering BAYES_80 or lower into a dedicated folder.
Sort by Subject, and go through the list. If the Subject alone isn't
sufficient evidence, have a quick look at the mail. Confirmed spam then
can be learned.
--
char *t="\10pse\0r\0dtu...@ghno\x4e\xc8\x79\xf4\xab\x51\x8a\x10\xf4\xf4\xc4";
main(){ char h,m=h=*t++,*x=t+2*h,c,i,l=*x,s=0; for (i=0;i<l;i++){ i%8? c<<=1:
(c=*++x); c&128 && (s+=h); if (!(h>>=1)||!t[s+h]){ putchar(t[s]);h=m;s=0; }}}
