On Wed, 15 Mar 2017, Kris Deugau wrote:
You'll also get users (mis)reporting legitimate mail of all kinds as spam (deliberately or otherwise); hand-sort the reports *before* feeding the messages to sa-learn. We have a few users that regularly report pretty much "all the mail in my Inbox", most of which is ham, and which would seriously break Bayes if we just passed those reports through unsorted.
Set those users up with per-user Bayes and train whatever they send. Negative feedback is a good engineering principle. -- John Hardin KA7OHZ http://www.impsec.org/~jhardin/ jhar...@impsec.org FALaholic #11174 pgpk -a jhar...@impsec.org key: 0xB8732E79 -- 2D8C 34F4 6411 F507 136C AF76 D822 E6E6 B873 2E79 ----------------------------------------------------------------------- W-w-w-w-w-where did he learn to n-n-negotiate like that? ----------------------------------------------------------------------- 449 days since the first successful real return to launch site (SpaceX)
