Gary Smith wrote:
I got an interesting email today that bayes was added to bayes as a
ham. The email was clearly a spam (which we have received many from
before). As you can see below it scored 0.0 on the basic test. It
received a bayes_56 scoring as can be seen below.
I find it interesting that a spam was THAT successful. The vast majority
of messages I get usually hit SOMETHING, so an actual spam dodging SA so
completely is interesting in and of itself. Do you have a copy you could
send? I'd be very interested in checking it out. Bayes has to train on
something first, and I generally count on the other rules to catch the
latest greatest variations.
It seems that the domain sayzsw.com seems to be clean in regards to
bayes and thus it levels out the score as can be observed below. Is
there a way that I can tweak bayes to reset or set a customized score
for an individual entry or to remove the entry entirely? Basically I
want to remove sayzsw.com from bayes so that this message would have
been considered spam and not relearned as ham in bayes.
Won't feeding that to sa-learn as --spam take care of that whole
process? Especially if the message scored that low overall! Or is there
some reason you don't want to re-learn it as spam?
- Bob
