Valery,
I think I know this one. Bayes will not auto learn unless the SA score is
signifigantly far enough away from your threshold. In your case, your letter
scored 4 with a threshold of 5. Being only 1 point away from the spam/ham
line, it will not be learned as either spam or ham. Now watch for a scored
letter with a large negative score or large positive score. These will have
"autolearn=yes" because the bayes routine is pretty sure they are spam or ham.
I can't remember what these autolearn thesholds are set at. Somebody else in
the know might mention it on the list.
scott
-----Original Message-----
From: Valery [mailto:[EMAIL PROTECTED]
Sent: Mon 3/1/2004 5:50 AM
To: [EMAIL PROTECTED]
Cc:
Subject: Autolearn problem
Hello!
I use SA 2.63 under FreeBSD 4.4.
There is record "bayes_auto_learn 1" in /etc/mail/spamassassin/local.cf
file
The field X-Spam-Status is given below.
X-Spam-Status: No, hits=4.0 required=5.0 tests=AWL,NIGERIAN_BODY1,
NIGERIAN_BODY2,NIGERIAN_BODY3,RISK_FREE autolearn=no version=2.63
date=Sun,
29 Feb 2004 20:52:25 +0300 bayes=0.5000
Why the autolearn=no is set to no?
What kind of mistake did I make?
Yours sincerely,
Valery
