Again and again..... The EPIM newsletter goes to spam. I marked it many many times as "not Spam"- but it doesn´t work. For some "Spam"emails it worked, but some are resistant to "not Spam".
Don´t ask me to send the newsletter, you should know it....
EPIM 9.7.5 portable. But it wasn´t working many versions before.
SPAM
Moderators: TerryRogers, Max
-
- Site Admin
- Posts: 15601
- Joined: Thu Nov 25, 2004 3:12 am
- Has thanked: 1411 times
- Been thanked: 984 times
Re: SPAM
We don't need the newsletter, we need to see the spam log file.
Android version of EssentialPIM. Keep all your data in sync!
-
- Site Admin
- Posts: 15601
- Joined: Thu Nov 25, 2004 3:12 am
- Has thanked: 1411 times
- Been thanked: 984 times
Re: SPAM
Thanks, according to the logs the spam filter reacted correctly based on its training data. So it all depends on how you trained it and whether you've been marking messages as spam and not as spam correctly and consequently. If you don't like the way spam filter works now, you can always reset its training data and then train it again carefully marking messages as spam and those which were detected incorrectly as not spam.
Android version of EssentialPIM. Keep all your data in sync!
Re: SPAM
Just for my understanding: Clicking Spam/not Spam is TRAINIG? I thought, to train a filter is on a statistic basis, that is, if I mark emails with the words "price", "cheap" and "buy" in it, than the filter tries to mark similar emails, containig similar words and perhaps email domains as SPAM".Thanks, according to the logs the spam filter reacted correctly based on its training data. So it all depends on how you trained it and whether you've been marking messages as spam and not as spam correctly and consequently. If you don't like the way spam filter works now, you can always reset its training data and then train it again carefully marking messages as spam and those which were detected incorrectly as not spam.
BUT if I mark your newsletter as NOT SPAM, than this newsletter, with the same sender and subject (EssentialPim) is marked as NOT SPAM. How can this definitly marked email go to spam again, by other training rules?
Re: SPAM
To a more precise understanding of this training: If the filter works on: my rules AND training data, and most of the emails are classified correct but some are resistant..... Should I reset training data or not? should I delete my rules? but I will have the same rules and training data again- so what´s the use of deleting it? Is there a way to work ONLY with rules without training?
-
- Site Admin
- Posts: 15601
- Joined: Thu Nov 25, 2004 3:12 am
- Has thanked: 1411 times
- Been thanked: 984 times
Re: SPAM
Spam filter doesn't take any message rules into consideration, it only uses its own Bayesian learning mechanisms.
Android version of EssentialPIM. Keep all your data in sync!
Re: SPAM
Then I would say, MANUAL RULES should overule Bayesian learned rules. Especially the "not Spam" MANUAL rule. Because, when I say, EPIM newsletter IS NOT SPAM- so there should be no way, the bayesian filter could this spam-excluded mail mark as spam again!
I think, the Bayesian filter should work only on NOT manually classiefied emails!
I think, the Bayesian filter should work only on NOT manually classiefied emails!
-
- Site Admin
- Posts: 15601
- Joined: Thu Nov 25, 2004 3:12 am
- Has thanked: 1411 times
- Been thanked: 984 times
Re: SPAM
There's an option for each of the mail rules to bypass spam filtering mechanisms for it.
Android version of EssentialPIM. Keep all your data in sync!