Hello Gunivortus, Monday, November 17, 2003, 12:53:46 PM, you wrote:
GG> Hi MikeD (3), >> It does for me, but what I had to do was to save a sample of "good >> mail" rather than just deleting them when I was done so that I had as >> many good messages (well, nearly so) as I did spam for the training. >> Once I did that I had no problems with "good" mail being classified as >> spam (false positive). GG> If there are enough good mails with the same address, which I marked as NOT GG> junk, I'd expect it working some next time; a mail with that address should stay GG> in the inbound. Well. I marked in the meantime about 40 times mails with the GG> same From-address, but still new ones are moved to the junkfolder. GG> I've had a correspondence with the maker of BayesIt, Alexey N. Vinogradov; I saw GG> them in the Junkfolder, marked them in TB under SPECIALS als NOT Junk, and moved GG> them back to my inbound. Nevertheless, another mail from him a week ago or so GG> was still moved to Junk. When you say that you moved it ... did you also "re-train" Bayesit? You do that by Right Click on the message to reclassify, select "Special" from that menu and the appropriate "Mark as Not Junk" or "Mark as Junk" from that list. You should see a "pop-up" saying that "1 message was processed". If you are not doing this, Bayesit does not know that you were unhappy with it's assessment. GG> A bigger problem is, I get f.i. an email from one of my readers, say GG> morrie45yahoo.com. I don't know him, it's the first, and maybe the last time I GG> get an email from him, but it's moved to the Junkfolder. Why? GG> I get lots of serious emails from people who maybe write me just one time, and I GG> answer just one time. All those incoming mails are seen as Spam. It is possible that you have your "threshold" set too low. You could try adjusting that. Also you need a lot of mail to train it, at least I did. I had to save 1000 spam and 1000 good messages to train it before I had good results. Make a folder (I called mine "not-spam") and instead of deleting those 'one off' good emails, move them there. I do that now with all my non spam and still save most of the received spam. When either folder gets to 2000 messages, I trim it back to 1000 by deleting the oldest messages. When I have done that I usually retrain using the more up to date messages. This works out to about once a month for me. Bayesit has been working very reliably for me since I started doing this. The biggest help was the message base of good messages. -- Best regards, MikeD mailto:[EMAIL PROTECTED] Using The Bat! v2.00.6 on Windows ME 4.90 Build 3000 ________________________________________________________ Current beta is 2.01.26 | "Using TBBETA" information: http://www.silverstones.com/thebat/TBUDLInfo.html
