Spamassassin and bayes learning

[koda]

I setup (following a guide for CPanel found on the web) a way to "feed" spam assassin Bayesan filters with legit and spam mails.
Now my doubt is ... when spam assassing finds a spam it tags it wit *** SPAM ... and so on... (I added the score too there) changing the original subject of the email.
So the mail I will have to feed the filters will include this modified subject line with "SPAM" in it and all the headers added by cpanel/spamassassin.
Will this mess up the learning process or bayesan filters just act on the actual BODY of the mail ignoring headers, sender, subject and everything else?
Thanks (sorry for being a noob)

 

[cPanelMichael]

Hello

I believe SpamAssassin considers the subjects of the message in addition to the message body. It's possible the manual method you are using to feed emails for learning could have a negative impact. Was the traditional method of having it learn over time not working to your satisfaction?

Thank you.

 

[koda]

Yes but that was "automated" tuning bayes filters based on what SpamAssassin already consider spam, but it's missing a lot of spam for italian native spammers.
In this other way I can "teach" him which mails are to be considered spam and which not, and this should be simplier (maybe more accurate) than defining new rules for italian language I think.
If it takes into account even subjects this may pose a problem ...
AT the moment I'm forwarding emails from multiple accounts to a single email address with 2 folders I use for this purprose one is for spam and one for ham. And then 1 time a day (or week) feed the SA bayesan filters.


EDIT: I'm starting to find something on this subject... for example here is a way to have bayesian filters ignore the SA headers:

bayes_ignore_header X-Spam-Flag
bayes_ignore_header X-Spam-Status
bayes_ignore_header X-Spam-Level
bayes_ignore_header X-Spam-Checker-Version
bayes_ignore_header X-Spam-Report
bayes_ignore_header X-Spam-Bar
bayes_ignore_header X-Spam-Score