On Wed, 20 Jan 2016 14:48:19 -0800
Marc Perkel <[email protected]> wrote:

> To be a little clearer. This new system isn't perfect. And it's main 
> strength is identifying good email. It does catch a lot more spam for 
> sure but when people scream at me it's because I blocked something 
> important. So think of this more as detecting ham as it's big feature.

I think you have the germ of a good idea in there somewhere.  It might be
that tweaking Bayes to emphasize highly-spammy and highly-hammy tokens
non-linearly (ie, giving them even more importance than the probabilities
would warrant while lowering the importance of "iffy" tokens) might
help, but it needs lots of experimentation and also I'd feel much happier
with some theoretical justification.

Regards,

Dianne.

Reply via email to