On Jan 29, 2007, at 7:39 PM, Frank Mitchell wrote:
But then why do spammers send messages full of random words? It seems
pointless to me.
The random words do help, to varying extents, against different types
of filters. And there's very little downside to including them.
On Jan 29, 2007, at 11:38 PM, Michael Lewis wrote:
I probably didn't communicate that well. As an example, this was in
my log:
Predicted: Good (27)
Trained: Good (Auto)
So, it was predicting this as good, and training it as good (auto). I
think that's the "Learning" function under Training preferences
kicking
in?
Yes, this is normal. The auto-training feature thought that this was
an interesting message because it was borderline (score of 27, with
50 being spam), so it decided to learn from it.
Then I'd click "Mark as Spam" and this would show up in the log:
Trained: Spam (Manual)
Mistake: False Negative
But this time it was wrong, so with your help it corrected the
training and recognized that it had made a mistake.
--
Michael Tsai <http://c-command.com>
