On Thu, 2005-03-03 at 17:19, Brian Mury wrote: > On Thu, 2005-03-03 at 13:55 -0800, Matt Florido wrote: > > "Finally, SpamAssassin requires at least 3 points from the header and 3 > > points from the body, to auto-learn as spam. If either section > > contributes fewer points, the message will not be auto-learned." > > I want SpamAssassin to be trained with all my mail. I tried to do this > by having SpamAssassin auto-learn from all mail, so I would only have to > retrain it with those messages that it gets wrong (which I can do with > Evolution's Junk/Not Junk buttons). This can't be done, due to what is > quoted above. I think there's a similar issue with auto-learning ham. The way this is handle usually is to manually run sa-learn on batches of ham and spam email. If this is done with a good sample in the first week or two of starting to use spamassassin you normally get a good base line and get very good results. Occasional training sessions will maintain things. About every two weeks I run a script that processes all spam I received. I run my inbox through less often. By manually doing this I get a chance to do a quick review looking for false positives. Anymore I don't spend much time on spam. And if you are running your own MTA server you can eliminate 99% of your spam before every having to invoke spamassassin by implementing greylisting. Seriously, it works that well. -- Scot L. Harris webid@xxxxxxxxxx "Yo baby yo baby yo." -- Eddie Murphy
