Matt Kettler wrote:
At 04:04 PM 3/18/2004, [EMAIL PROTECTED] wrote:
Two questions:
a) is there more documentation somewhere regarding using "bayes_ignore_header"? Are there any default settings at all? What about the "To:" header? Do we need to explicitly ignore that (do we want to)?
The list of default ignores is way too long to post here.. Look in Bayes.pm for $IGNORED_HDRS.
To: is not on the list, but Resent-To: is.
I don't think you want to explicitly ignore the To: header.. it's a very good token for training SA about mailing lists, etc, and ignoring it could dramaticaly reduce the accuracy of your bayes database.
b) is there an "ignore" type config for sa-learn?
Um.. yeah.. bayes_ignore_header.... sa-learn uses the same config files and options as spamassassin. There's no separate options for sa-learn.
Well, given the asymmetric situation in the "b" part of my mail, if I'm not going to ignore the To: header when scoring, I'd like to ignore it when learning. Otherwise the people who save spam for use as a learning corpus (i.e. me :-) will have the To: my-address scored too highly negative and not balanced in the auto-learned ham.
Has anyone else run into this issue? Or am I missing something and it's not really an issue?
-glenn
