On Jul 8, 2010, at 11:35 AM, [email protected] wrote:
from what you've said it's hard to know why it's not recognizing some of your incoming mail as spam (though Vince has a couple of good hunches) but again, spam recognition is not based on the sender's address, it's based on "latent semantic analysis", which is a clever and trainable but fallible filtering technique that attempts to discern spammy "meanings" in email; it is very complex, you can look it up if you want to delve into the theory, but the upshot is that there will be no "log" that says what was filtered and why -- the info is kept a database and the Reset button literally resets that database to its default state, and you then try to train it and see how it goes; in my experience, the fancy theory does not yield superb results
I do not know how the OS X learning system works etc. However, I use spam assassin on my servers and it has a similar system and for us it works very well. Very few spams escape its clutches and it makes very few false positives. However, when you start fresh with spam assassin the learning engine needs to be fed enough that it can start to discern things. A few spams fed into it does not usually work and it won't actually start working until you have trained it sufficiently. Maybe the Apple one is similar? I do not know, but am comparing to a similar system. Maybe clicking JUNK on enough mail will trigger the system to start working.
Chad _______________________________________________ MacOSX-talk mailing list [email protected] http://www.omnigroup.com/mailman/listinfo/macosx-talk
