[Spambayes] Sample spam clues for white-on-white message

Gray Norton gray at stanfordalumni.org
Tue Oct 28 14:49:01 EST 2003


> -----Original Message-----
> From: Ryan Malayter [mailto:rmalayter at bai.org]
> Sent: Tuesday, October 28, 2003 10:45 AM
> To: Gray Norton; spambayes at python.org
> Subject: RE: [Spambayes] Sample spam clues for white-on-white message
> 
> It sure looks like a copy of this message was trained as ham before
you
> ran the spam clues, since so many of the tokens unique to this message
> are listed with one ham count.

I agree -- that definitely appears to be the case.

> Are you sure you didn't mis-click a training button? Do you have the
> plug-in set up to train when you move messages from folder to folder?
(I
> cannot remember if SB does this by default.) Could you have moved it
by
> accident? What about an Outlook rule moving the message?

I'm positive I didn't mis-click, so I'm thinking that the message-moving
theory is the one to investigate, especially since I do have Outlook
rules that move all of my messages when they come in.

Looking at my filtering setup, I see that I have filtering enabled for
my main Inbox folder, as well as for all of its subfolders, into which
my Outlook rules move messages. I set things up this way because I
wasn't sure how Spambayes interacted with Outlook rules from a timing
standpoint and I figured it couldn't hurt to check all of the folders.

I suppose it's possible that some or all of my spam gets filtered from
the Inbox into the Junk E-Mail folder, then somehow snatched back out of
Junk E-Mail by an Outlook rule and deposited in one of my Inbox
subfolders, where it gets checked again -- this would account for the
scoring I'm seeing. If so, however, it is happening quickly enough that
the Outlook GUI doesn't show it.

For now, I'm going to 1) Disable the train-on-move options; and 2) Stop
filtering the main Inbox folder and filter only the subfolders instead.
We'll see whether that solves the problem.

> I turn the "train when messages moved" features of SpamBayes off, so I
> only get training when I pusposely click on the "delete as spam" or
> "recover as ham" buttons.

So, if I understand you correctly, the train-on-move options displayed
in the Spambayes Manager apply when messages are moved by drag-and-drop
or by Outlook rules, but NOT when messages are moved by Spambayes'
"delete as spam" and "recover as ham" buttons (in which case training
happens regardless of the setting). Have I inferred correctly? If so,
then I was definitely mistaken in my original interpretation of these
options.

Thanks for your help,

Gray  






More information about the Spambayes mailing list