Agreed.  Setting the bayes threshold is an idea.  I will play around
with it on the development box and see how it works out for me.
Unfortunately I no longer have a large spam collection to retrain from.
When I setup the bayes I had ~5000 spams that I started with as a
baseline but I think I removed then mboxes when I meant to copy them.
DOH!

>just don't see much sense in the default of "learn as ham if it doesn't
>contain any known spam markers". My philiosophy is that learning as ham
>needs to be earned, not just a default.

Gary Smith

Reply via email to