-------- Original-Nachricht --------
> Datum: Wed, 12 Sep 2007 17:53:08 -0700
> Von: Tom Bombadil <[EMAIL PROTECTED]>
> An: [email protected]
> Betreff: [dspam-users] using --source=error without a signature

> Is there anyway to train a message when the signature is not present?
> 
> basically I want to use:
> dspam --class=spam --source=error
> 
--source=error can only be used if DSPAM can reference the (error) tokens. So 
this only works with a previously processed message and a valid signature.

If you want to train without signature, then you need to use another source 
then error. So only corpus or inoculation is left:
dspam --class=spam --source=corpus
dspam --class=spam --source=inoculation

The problem with corpus or inoculation is that they are not designed for 
correcting misclassification. At least they don't offer a way to switch tokens 
from one class to the other (what error does). You could however train with 
corpus/inoculation in a loop for the correct class until the message processed 
results with the desired class.

I would use corpus for forced training and inoculation for spam inbound on 
honeypots and/or for ham on outbound mail from your users.


> If the signature is not present, the command above fails.
> 
Yes. Error needs a reference and this reference is the DSPAM signature. Without 
the DSPAM signature you can not use error.


> Cheers,
> g.
>
cheers

Steve
-- 
GMX FreeMail: 1 GB Postfach, 5 E-Mail-Adressen, 10 Free SMS.
Alle Infos und kostenlose Anmeldung: http://www.gmx.net/de/go/freemail

Reply via email to