On Thu, 2022-09-01 at 22:12 +0930, Tim via users wrote:
My understanding was that the interpretation had already been done, and they remain in the junk folder in case it got it wrong. You can check what it's junking, you can find a mail it erroneously junked, and you can reclassify it as being non-junk.
For what it's worth, that's certainly the way other programs worked. As mail is processed, spam rules are made somewhere else.
Back when I used a program using Bayesian rules, I quickly programmed it in two stages: Using a folder of lots of kept spam, selected them all, and flagged them as spam. And folders of kept non-spam, selected them all, and flagged them as non-spam.
It was quite effective, but for one thing: If a mailing list became a source of spam, and you kept marking mail from it as being spam, the system can add the mailing list itself as being spam.
Eventually I settled on not using junk detection, using a separate email address for mailing lists, having that address *only* accept mail from those lists. You get no false positives (real mail getting falsely flagged as spam), and the few that slip through I just hit the delete button once or twice each day (yes, it's that few).
False positives are a huge curse. You lose touch with friends, you miss jobs and incoming bills. If you have to keep checking your junk mail folder for them, what's the point of using spam filtering? You're going to see the spam, anyway.