[Python-Dev] Tracker marks my messages as spam :-)

skip at pobox.com skip at pobox.com
Fri Feb 1 20:48:37 CET 2008


    Guido> """  
    Guido> Oracle confirms the issue. They will provide a patch.
    Guido> """

    Guido> That's just small, but has a high S/N ratio. The contents of Jesus'
    Guido> email has nothing to do with this issue.

As Martin pointed out, small messages tend to get classified as either spam
or unsure.  The spam filter built into the Roundup instance uses the
SpamBayes classifier.  I don't know how many examples have been trained so
far, but I would guess very few.  It's unlikely that the small message gave
any useful clues (far enough away from a score of 0.5 in either the spam or
ham directions) to the classifier.  Maybe "Oracle" or "patch" would have
been hammy clues.  The others were probably tossed out.  In short, there
just wasn't enough "meat" to chew on.  Of course, without seeing the
classifier's database and input it's kind of hard to be more precise.  Over
time though, even such short messages should be classified more accurately
as the training database grows.

Skip



More information about the Python-Dev mailing list