-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Wednesday, 2009-09-09 at 23:02 +0200, Roger Oberholtzer wrote:
On Mon, 2009-09-07 at 21:09 +0200, Carlos E. R. wrote:
sa-learn --spam --showdots --dir /pathto/spam
(don't forget to feed it both spam and ham mail, the results are much better)
I just noticed that my SPAM info in the message (put there by SA) is:
tests=BAYES_99,HTML_IMAGE_ONLY_04, HTML_MESSAGE,HTML_SHORT_LINK_IMG_1,MIME_HTML_ONLY,TVD_SPACE_RATIO autolearn=no
When you see the "BAYES_" there, then it does bayes testing. It is one of the few places where you can know.
The autolearn item is odd. I am guessing this is not referring to sa-learn.
Right :-)
If not that, how would SA do autolearn? If it does not get it right the first time, short of telling it when it was wrong, how would it learn?
Or, does this mean that whatever it has learned in sa-learn will not be used? Is learning one thing, and using it another? If so, software gets more and more like people every day!
Simple: it means that it has (not) added that email to the bayes database; as spam if it was marked such, or a not spam if not. Ie, it learns on each email it scans as it is received. However, if it doesn't detect mail correctly, and that email is added to the database, then the bayes tests are further biased wrong each time. I prefer to disabled autolearn completely and do manual learning instead. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.9 (GNU/Linux) iEYEARECAAYFAkqoKMoACgkQtTMYHG2NR9W5igCfXLmjwSjJS827+/q/LmVRNNAc uzIAnAk3izRl0150+QhqkINVt/0A5g4a =KrQm -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org