According to your logfile, you've trained "4 spam and 412 good" messages. This is almost certainly the reason for your problem.
 
This is an *extreme* oversimplification of the mathematics in SpamBayes, but the SpamBayes filter basically boils down to taking each word in the message and comparing the percentage of spam messages that contain the word to the percentage of good messages that contain the word. Let's say you have a word that appears in 2 spam messages. That's 50% (2/4). To get an equal 50% of good messages to balance the spam percentage, the word would have to appear in 206 good messages (206/412 = 50%). As you can see, it doesn't take nearly as many occurrences of a word to make it seem spammy as it does to make it seem good.
 
To solve the problem, I would start by resetting your training data and starting over from scratch. To do that, first close Outlook and then go to your "\Documents and Settings\username\Application Data\SpamBayes" directory and delete the *.db files. Now restart Outlook. With no training data, any new messages should go to your "Unsure" or "Possible Spam" folder. Select each unsure message and select either "Delete as Spam" or "Recover from Spam" to train SpamBayes with the correct classification. With a little training, SpamBayes will soon start identifying new messages as either good or spam. Just keep an eye on the results and use the toolbar buttons to train any messages that SpamBayes classifies incorrectly or as unsure. You should find that the accuracy of the SpamBayes filter will increase very quickly.
 
You can read more about training on the SpamBayes wiki. This particular style of training is referred to as TrainOnErrorsAndUnsures on the wiki.
 
 
--
Kenny Pitt
 


From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
Sent: Tuesday, May 24, 2005 3:47 PM
To: [EMAIL PROTECTED]
Subject: RE: [Spambayes] Spam Help

Kenny,

Thanks for the tip. Here is the log file.

There is no server side filter too.

 


From: Kenny Pitt [mailto:[EMAIL PROTECTED]
Sent: Tuesday, May 24, 2005 3:40 PM
To: Darrell Rogers; spambayes@python.org
Subject: RE: [Spambayes] Spam Help

 

If SpamBayes is running and moving messages to the Spam folder then it should also be creating a log file. Try searching your hard disk for any file named "spambayes?.log" (where ? could be any of the digits 1, 2, 3, or 4). The log file would be very helpful in determining what is causing this.

 

You could also use the "Show spam clues" command in the SpamBayes menu and send us the clues for one of the messages that was incorrectly moved to the Spam folder. The clues will usually allow us to determine if there is a training problem that is causing the misclassification.

 

If there really is no log file on your disk then it is highly unlikely that it is SpamBayes that is moving the messages. In that case, you should check if there is another spam filter running or if you have any message rules that might be moving the messages. If you are connecting to an Exchange server, there is also a slight possibility that your organization might be running a server-side spam filter that is moving the messages before you access them.

 

--

Kenny Pitt

 

 


From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of [EMAIL PROTECTED]
Sent: Tuesday, May 24, 2005 10:45 AM
To: spambayes@python.org
Subject: [Spambayes] Spam Help

Hello,

All of my email is being filtered as spam. I goes directly to the Spam folder. I checked you trouble shooting guide, but still could not fix. I even upgraded to your latest version of SpamBayes...1.0.4 (March 2005) Can you help?

Thank you.

 

Sincerely,

Darrell Rogers

 

Op System isWindows 2000 running Outlook 2000

SpamBayes 1.0.4 (March 2005)

There was no log file

 

_______________________________________________
Spambayes@python.org
http://mail.python.org/mailman/listinfo/spambayes
Check the FAQ before asking: http://spambayes.sf.net/faq.html

Reply via email to