3.0.2 - will not learn good

You already have the messages in question in the Junk mailbox in Mail. So you could just pick a couple of messages and search for them (by sender or subject) in SpamSieve’s log.

Or drag the scroll elevator to quickly get to the right date range, since you know when you installed the update.

The video helps, thanks. It’s not necessary to count or scroll through them all.

It looks like some of these really were spam, but others at the start were not. SpamSieve got some of those wrong and auto-trained itself that they were spam. Since this was a large batch of messages, with no time for you to correct it in the middle, this incorrect auto-training made SpamSieve think a lot of the later messages were spam, too. So I think it’s now clear what happened.

Going forward:

  • You can see in the log that most of the messages are still shown with green circles. That means SpamSieve thinks it got them right—that they really are spam. In order to have good filtering accuracy in the future, you should correct all the mistakes or choose Reset Corpus from the File menu, then re-train SpamSieve.

  • I will look into adding a safety feature so that old messages are not processed when you select a mailbox for filtering, even if they are unread.

Correct.

At this time, SpamSieve is correctly identifying Spam, and not producing false positives. “MY ISSUE” is that Google’s gmail was trained on the bulk transfer, and is creating the false positives not SpamSieve.

I will work on using SpamSieve to ‘train as good’ in the junk folder to move the right ones back.

At this point, I do not think there is a SpamSieve ‘issue’ (some tuning, but no issue) the issue is that gmail is aggressive on spam filter, as, for an example USPS email that I get 6 days a week, had 900+ of it’s emails moved to JUNK, the gmail spam filter is using that as the reason to filter the emails. I need to un-train gmail.

Your thoughts on using SpamSieve to move them is not working

  • Email from CVS, ORANGE BACKGROUND
  • Train as good, in the log as such
Summary:	You trained this message as Good from your mail client.
Prediction:	SpamSieve had not previously seen this message. This could be because (a) you were doing an initial training, (b) you had marked the message as read on another device, (c) your mail program was not set up to filter all new messages through SpamSieve, or (d) a server filter had moved the message out of the inbox before it got to your Mac.
Help:	Do an Initial Training
Help:	How should I configure the junk filter on my mail server?
Help:	Why do good messages keep going to the Junk mailbox?
Corpus Change:	Message was already in Good corpus (1,022 total)
Repeated Training:	You had already trained this message.
Date Logged:	Today, 10:46:05 AM

Subject:	CVS Caremark Alert: Order Received Confirmation
  • moved to inbox, get moved back to junk

gmail is moving things back

Why is this message in spam?

It is similar to messages that were identified as spam in the past.

as a note. Durring the ‘bulk’ move there were 66 logs of CVS moved to JUNK

As seen above, spamsieve knows it is not spam, but Google is moving it.

  • in JUNK
  • click on email, click train as good, get the above log
  • email pops out of Junk, then pops back into Junk
  • clip of why Gmail is doing this.
  • Use google web to unmake as spam, moves and stays in inbox

That’s strange, as I’ve consistently seen that Gmail’s spam filter doesn’t reprocess old messages. And, even if it did, having SpamSieve tell Mail that the message is not junk and move it to the inbox should be equivalent to marking the message as Not Spam in Gmail. Usually what you are reporting happens because of a mail rule on another Mac or PC. So I’m afraid I don’t know of another solution, and it might be worth asking Google. There’s some more discussion of this here.

There’s a change in SpamSieve 3.0.3b1 to prevent automatic filtering of old messages.

1 Like