On Tue, 08 Jun 2004 10:06:56 -0400 "Matt Kettler" <[EMAIL PROTECTED]> wrote:

> No, it wasn't skipped. There's no BAYES_ rule that matches when the bayes 
> score is exactly 50/50. Take a look at the bayes rules, there's a tiny gap 
> with no rule matching between 0.4999 and 0.5001
> Generaly this means the message didn't contain any tokens or very few 
> tokens that were in your bayes DB.
> You can reduce the occurance of this with more training, but it will still 
> happen once in a while.
> For reference, SA 3.0 appears to have a BAYES_ rule that matches at exactly 
> 50%, probably to keep these questions from cropping up.

understood! thanks for the clarification.

Reply via email to