Larry Rosenman skrev den 2014-01-15 00:09:
On 2014-01-14 17:07, Benny Pedersen wrote:
is it possible to learn it as ham with bayes ?, bayes 50 is more or
less neotral score for unsure content, help learning it as ham will
solve it to go under 5
Already done, but this seems a bit high for a first time hit :(
if bayes 00, then it would be under 5
i use sagrey to first time senders here, no false hits on it
but make a local rule to help on it to turn it down under 5, base it on
content, not ip/sender domain, but content you like to be ham, keep
scores of max 1.5 and -1.5
sa scores is generated from spam samples that turn this score high, and
it could mean there is to few ham samples for this kind of mails :(