The spam phrases stuff is calculated statistically based on a largish corpus
of spam and nonspam emails (close to 100,000 messages all together, about
half of each).  You can find (slightly) more details in the wordfreqs
directory in CVS -- basically there's a batch run which counts the frequency
of occurrence of phrases in spam mails vs. nonspam mails, then that file is
generated automatically.  You're not going to be able to calculate spam
phrase lines by hand.  Best to just create a new rule.  I'm planning on
taking a closer look at spam phrases later, since I don't fully understand
it myself.

C

on 2/20/02 7:21 AM, Mike Grau at [EMAIL PROTECTED] wrote:

> Hello
> 
> I am running SpamAssassin as a milter and am
> very pleased indeed.
> 
> Can someone give me a brief explaination of the
> 40_spam_phrases.cf contents? For example, as in
> "spamphrase 29530 seventh heaven" what are the
> scores in the second column and how are they
> determined? Is it common to add your own spamphrases
> in your local configuration and if so, how do
> you determine the number you give it?
> 
> Or is it more common to just add a body test and not fool
> with spamphrases?
> 
> Thanks - Mike


_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to