On 06-Mar-22 20:25, Felipe Gasper wrote:

On Mar 6, 2022, at 20:11, Ricardo Signes<perl.test...@rjbs.manxome.org>  wrote:

On Thu, Mar 3, 2022, at 09:48, Felipe Gasper wrote:
   1.0 BAYES_999              BODY: Bayes spam probability is 99.9 to 100%
                              [score: 1.0000]
   5.0 BAYES_99               BODY: Bayes spam probability is 99 to 100%
                              [score: 1.0000]
These relate to your training of SpamAssassin.  I don't know how you or your 
provider is training your Bayes db, but you're getting six points from that.
It’s a cPanel server.

Maybe the `nbsp` in there--useless in a plain-text email?--is the culprit. 
(Bayes, from what I’ve read, takes into account misspelled words.)

   3.3 EXCUSE_REMOVE          BODY: Talks about how to be removed from mailings
This is, I think, a bit higher than the default value for this, but not much higher.  I 
think that's too high, but perhaps replacing this with "to unsubscribe" would 
help.  (Again, though, your personal Bayes is the likely problem here.)
The list-help, list-unsubscribe, etc. headers would probably be a good addition 
here?

   0.9 PP_MIME_FAKE_ASCII_TEXT BODY: MIME text/plain claims to be ASCII
                               but isn't
The email does not have a Content-Type header.  It should have one, probably 
"text/plain; charset=utf-8"
Agreed.

-F

The simplest solution would be to feed these reports into the bayes filter.  sa-learn is the utility for that.

E.g. file them in a folder, and do something like

sa-learn --ham --mbox ham_folder

I keep training folders for spam, ham, and forget & have a daily cron job to keep the filter up-to-date.

sa-learn has a bunch of options for tuning its behavior, which man will turn up.


Attachment: OpenPGP_signature
Description: OpenPGP digital signature

Reply via email to