Tips on Manual Bayes Training?

[Nathan Johanson]

Quick question for those of you with well-trained bayes databases.

I'm planning to set up some spam traps. Question: Is there any advantage
to learning messages already marked as spam by SpamAssassin?
Logistically, it makes sense only to feed false negatives and false
positives.

For the time being, I'm planning on using MailScanner's "Non Spam
Actions" ruleset to forward unmarked spam sent to (postmaster@, info@,
sales@, etc.) to a spamtrap mailbox. I'll verify all messages as false
negatives and then learn them into the bayes database. This is an
attempt to offset some of the poisoning that's been affecting us lately.
This doesn't take ham into account, but then I haven't had a lot of
problems with false positives. 

Any suggestions or alternative methods? 

I like the idea of end users redirecting spam to the appropriate
spam/ham mailboxes, but the majority of them are using Outlook or
Outlook Express and don't have any way to do this.

Nathan