One other detail that might be nice, Anthony, is stopping processing
email while sa-learn runs. I run it at normal priority once a night.
I also feed BAYES a starvation diet based on manual (only) training.
(Hey, you can DO things like that for your own setup. If you are running
an ISP you MUST make compromises between users who can't make the 1950's
toaster power cord work on their new computer when they threw the new
computer's power cord away with the packing materials and Bayes accuracy.)
I've trained on less than 2000 messages each for ham and spam. If a
message has meat of any kind and was missed by Bayes it gets used for
training. Otherwise - why waste CPU cycles even if the CPU is already
idle.
{^_^}
----- Original Message -----
From: "Anthony Messina" <amessina@xxxxxxxxxxxx>
<I'd have quoted something here if Anthony had really used plain text
with no mime <excrement>.>
