Somebody already answered the syntax for modifying your scores in your local.cf, so...

The auto-learn "bayes evaluator" doesn't take the Bayes scores into account when 
deciding whether to learn as spam or ham.  So you could have autolearn threshold set 
to 10, have your Bayes tests at 20 points, and get a 29-point spam come in, but not be 
sent through autolearn. 

> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> Sent: Friday, August 29, 2003 11:43 AM
> To: [EMAIL PROTECTED]
> Subject: [SAtalk] Changing Bayes scoring
> 
> 
> Greetings,
> 
> I'd like to increase the score for certain bayes
> confidence levels.  My understanding is that I
> need to put one (or more) of these lines from 
> /usr/locals/hare/spamassassin/23_bayes.cf:
> 
> 
>         body BAYES_60           eval:check_bayes('0.60', '0.70')
>         body BAYES_70           eval:check_bayes('0.70', '0.80')
>         body BAYES_80           eval:check_bayes('0.80', '0.90')
>         body BAYES_90           eval:check_bayes('0.90', '0.99')
>         body BAYES_99           eval:check_bayes('0.99', '1.00')
> 
> into /etc/mail/spamassassin/local.cf
> 
> and in order to increase the scores from certain bayes confidence
> levels, make edits, such as:
> 
>         body BAYES_70           eval:check_bayes('0.70', '1.80')
>         body BAYES_80           eval:check_bayes('0.80', '1.90')
>         body BAYES_90           eval:check_bayes('0.90', '2.99')
>         body BAYES_99           eval:check_bayes('0.99', '3.00')
> 
> Or am I off the mark entirely?
> 
> A good 80-90% of the spam that scores 4.3-4.9, while being all over
> the place wrt other scores (i.e. some have bad mime, some have bad 
> html, some have bad times), they have bayes confidences of 70-99.
> As well the few lists that I'm on seem to get 3.5 to 4.5-ish scores,
> mostly for the bad html that comes from their MUA, yet have pretty 
> consistent bayes confidences around 20-50%.  
> 
> So, it seems to me that I should rely a bit more on bayes, as just
> lowering the threshold will get a fair bit of ham.
> 
> Yet, in doing so, does this feed back into the bayes evaluator?
> 
> Cheers!
> -sam
> 
> 
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Spamassassin-talk mailing list
> [EMAIL PROTECTED]
> https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
> 


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to