You can add rules, but you'll increases the chance of false positives. Auto-learning is designed to only work on the spammiest spam, that fires on many rules. It then trains Bayes to find similar spam that might be evading those rules. So, I'd recommend not changing anything, since it seems like SA is working just fine.
http://wiki.apache.org/spamassassin/FalsePositives http://wiki.apache.org/spamassassin/AutolearningNotWorking - dan -- Dan Kohn <mailto:[EMAIL PROTECTED]> <http://www.dankohn.com/> <tel:+1-650-327-2600> -----Original Message----- From: David Jim�nez Dom�nguez [mailto:[EMAIL PROTECTED] Sent: Monday, July 12, 2004 14:45 To: [EMAIL PROTECTED] Subject: Re: autolearn problem > David Jim�nez Dom�nguez wrote: > > >Hi lists!! > > > >I'm using Amavis-New with SA and Postfix. When I check the Spam in the > >quarantine directory, all the messages doesn't have the autolearn label in > the > >headers (Neither autolearn=no, nor autolearn=SPAM / HAM). This is my > local.cf: > > > >bayes_auto_learn 1 > >bayes_auto_learn_threshold_nonspam 0.1 > >bayes_auto_learn_threshold_spam 10.0 > > > > > > > > >and this is an Spam header... > > > >X-Spam-Status: Yes, hits=11.8 tag1=-999.0 tag2=3.0 kill=5.0 > tests=ALL_NATURAL, > > BANG_GUARANTEE, BAYES_99, HTML_FONTCOLOR_RED, HTML_FONTCOLOR_UNSAFE, > > HTML_FONT_BIG, HTML_MESSAGE, MIME_HTML_ONLY, MONEY_BACK > >X-Spam-Level: *********** > > > >What is wrong??? Please help me > > > > > > The pre-Bayes score is not over 10.0, which is your threshold for > autolearning. Without parsing the individual scores, keep in mind you > have to have a certain number of points from header and body tests also > - it looks like you have 0.0 from headers. > > -- > -tom > I read about pre-bayes test from you before in this list, Could I grow the header and body scores in the pre-bayes test, adding custom rules like SARE?? What can I do, to grow the score in the pre-bayes test???
