Spam Folder: Difference between revisions

Back up to the Email Features Category
From AAISP Support Site
mNo edit summary
Line 28: Line 28:
If the 'Bayes spam probability' is incorrect, then it can be taught about messages that it has got wrong in the following ways:
If the 'Bayes spam probability' is incorrect, then it can be taught about messages that it has got wrong in the following ways:


*Move a message '''OUT''' of the spam folder to tell the system it is '''NOT''' spam
*Move a message '''OUT''' of the spam folder to tell the system it is '''NOT spam'''
*Move a message '''IN''' to the spam folder to tell the system it '''IS''' spam
*Move a message '''IN''' to the spam folder to tell the system it '''IS spam'''


If you make a mistake, just move the message back.
If you make a mistake, just move the message back.

Revision as of 11:39, 19 August 2019


When connecting to our email service over IMAP we will create a special folder named spam (lowercase). If you delete the spam folder it will be re-created next time you log in.

The service will then move messages that has been classified as spam in to this folder. This helps keep your inbox free of spam and you should keep an eye on this folder in case there are messages that are not spam.

Disabling the spam folder

To stop the mail system from putting spam (or any) messages then you can create a Whitelist for your address on the Control Pages.

Spam 'learning'

Our spam checking system performs many checks on a message to decide whether it is spam or not. One of the checks it can do, is to compare the message with previous messages -this is called Bayesian Learning.

Checking the learning

If you view the 'headers' of a message, you should see the spam report, and a line looking like:

 -3.0 BAYES_00               BODY: Bayes spam probability is 0 to 1%

or perhaps

 0.8 BAYES_50               BODY: Bayes spam probability is 40 to 60%

Here you can see the 'Bayes spam probability' as a percentage. In the first example the Bayes really thinks the message is not spam - it says 0-1%. The second example is less sure about it.

Teaching the system

If the 'Bayes spam probability' is incorrect, then it can be taught about messages that it has got wrong in the following ways:

  • Move a message OUT of the spam folder to tell the system it is NOT spam
  • Move a message IN to the spam folder to tell the system it IS spam

If you make a mistake, just move the message back.

The process of learning doesn't guarantee that the system will be correct next time a similar message is received, but it should help and you should see a change in the 'probability' in the headers.

Legacy folders

In the past we used to have a system that used folders called learnham and learnspam - this system is no longer used. These folders can be removed.