Hello Tech,

Thursday, June 3, 2004, 11:34:35 AM, you wrote:

T> --- Gary Smith <[EMAIL PROTECTED]> wrote:
>> It's always good to train it with something.  It
>> will learn on it's own as well.  We do both.

T> OK...So this means I should run the sa-learn command
T> right?

Yes, I agree.  auto-learn is ok. It's more than OK -- it's very useful
and beneficial.  But hand-feeding it manually verified ham and spam is
even better.

I use auto-learn to keep Bayes up to date moment to moment.  I then on a
regular basis do sa-learn with a collected spam file and collected ham
file to feed it more samples/information.

"Regular basis" here means "whenever I have a false positive" (haven't
had a false negative in months), and otherwise twice a week. Your
schedule might differ according to amount of emails.

Bob Menschel



Reply via email to