Hi guys!
I've found some strangeness in my battle against spam with dspam. :)
I'm using a postfix setup, and use the ParseToHeaders thingy to just be able to
send to [EMAIL PROTECTED] or [EMAIL PROTECTED] to relearn spams and hams that
were misclassified. That way i don't need to use the dspam-add and dspam-del
services in postfix' master.cf (iirc).
However, this is my dspam_stats output:
@verwilst.be:
TP True Positives: 18
TN True Negatives: 8
FP False Positives: 8
FN False Negatives: 1
SC Spam Corpusfed: 0
NC Nonspam Corpusfed: 32
TL Training Left: 2452
SHR Spam Hit Rate 94.74%
HSR Ham Strike Rate: 50.00%
OCA Overall Accuracy: 74.29%
How does he get the NC values? I never corpusfeed? I always forward to notspam@
... It did notice a FP correctly once though, while i'm sending to the same
address?
Any clues?
Also, if i do
/usr/bin/dspam --mode=teft --user @verwilst.be --class=innocent --source=error
--debug < mail.txt
Shouldn't it at least do ANYTHING ? :) It just doesn't seem to do anything. I
have my uid in the signature, so it should just find the correct user from
there, and relearn it, right? It doesn't change a thing to my stats or
system.log though..
Thanks a lot!
Bart Verwilst
