On Thu, 2004-02-26 at 17:36, Kelson Vibber wrote: > At 07:27 AM 2/26/2004, G. Roderick Singleton wrote: > >Within the past few days, someone published that he had put together files > >that could be used with sa-learn. > > Any Bayesian classifier will be most effective when trained on the > spam/non-spam you actually receive. Using someone else's mail will always > be less accurate than using your own. > > Just save your own spam and non-spam collections to a pair of mbox folders > and run sa-learn <mode> --mbox on each. >
Thanks. I just wanted to see whether or not the canned solution was effective. I know I saw the message, I even had the web site in my browser but it crashed and I lost the info and now I can't find the original message. Now to do what you suggest will require moving a mailbox over the wire to the mailhost and I have no general "ham" with which to seed easily. And since I need to retrain I had hoped to try the prepared packages. ger _______________________________________________ Visit http://www.mimedefang.org and http://www.canit.ca MIMEDefang mailing list [EMAIL PROTECTED] http://lists.roaringpenguin.com/mailman/listinfo/mimedefang
