Reporting ham (non-spam), Bayesnian filtering

eewanco · 23 Oct 2015, 04:58 AM

I am having spam problems but I'll report on that once I've exhausted all avenues. One of those avenues is individual Bayesnian filtering, which I've discovered is not enabled because I haven't reported enough ham (non-spam). According to Advanced/Spam Virus Protection, "Each user has an individually trainable bayes database to improve spam filtering. You can train your personal bayes database by selecting some messages and using the Report spam and Report Non-Spam options from the Action menu on the Mailbox screen." (This page also indicates that I haven't reported enough ham, and that I'm using the global database rather than an individual one.) I've looked all over; I can find the Report spam button, but not the Report Non-Spam button. I can set the folder as non-spam, but I'm not sure this folder is purely non-spam and I don't want to mess up the training. I'd rather use the button, though. Where is it, or is it gone?

Also how does the folder training work -- is it cumulative? In other words if I mark my inbox as "ham" and some spam gets through overnight just before it scans it, how badly will this throw off the results over time?

janusz · 23 Oct 2015, 04:20 PM

Quote:

Originally Posted by eewanco

I can find the Report spam button, but not the Report Non-Spam button.

It is there when you open the Spam folder. Predictably, the Report spam button is absent when viewing that folder.

n5bb · 24 Oct 2015, 01:17 PM

Also remember that each folder can be set to learn messages as spam or non-spam in the Advanced>Folders screen. Only messages which are new in that specific folder from one day to the next (during the daily folder scan process) are learned.

So you can create a new folder and set it to learn messages as non-spam.
Then move ham messages to that new folder. Leave them there at least 24 hours.
You can then move them to any folder, since they are now learned as ham.
I wouldn't worry about the timing. The last learning action taken on a message is what matters. If spam arrives at your Inbox and you don't mark it as non-spam and your Inbox is set for learning as non-spam, then those messages will temporarily affect your spam filing. But as soon as you mark them as spam your user Bayes filter will be fixed.

Bill

23 Oct 2015, 04:58 AM	#1
eewanco Junior Member Join Date: May 2008 Location: Shrewsbury, MA USA Posts: 13	Reporting ham (non-spam), Bayesnian filtering I am having spam problems but I'll report on that once I've exhausted all avenues. One of those avenues is individual Bayesnian filtering, which I've discovered is not enabled because I haven't reported enough ham (non-spam). According to Advanced/Spam Virus Protection, "Each user has an individually trainable bayes database to improve spam filtering. You can train your personal bayes database by selecting some messages and using the Report spam and Report Non-Spam options from the Action menu on the Mailbox screen." (This page also indicates that I haven't reported enough ham, and that I'm using the global database rather than an individual one.) I've looked all over; I can find the Report spam button, but not the Report Non-Spam button. I can set the folder as non-spam, but I'm not sure this folder is purely non-spam and I don't want to mess up the training. I'd rather use the button, though. Where is it, or is it gone? Also how does the folder training work -- is it cumulative? In other words if I mark my inbox as "ham" and some spam gets through overnight just before it scans it, how badly will this throw off the results over time?

24 Oct 2015, 01:17 PM	#3
n5bb Intergalactic Postmaster Join Date: May 2004 Location: Irving, Texas Posts: 8,930	Also remember that each folder can be set to learn messages as spam or non-spam in the Advanced>Folders screen. Only messages which are new in that specific folder from one day to the next (during the daily folder scan process) are learned. So you can create a new folder and set it to learn messages as non-spam. Then move ham messages to that new folder. Leave them there at least 24 hours. You can then move them to any folder, since they are now learned as ham. I wouldn't worry about the timing. The last learning action taken on a message is what matters. If spam arrives at your Inbox and you don't mark it as non-spam and your Inbox is set for learning as non-spam, then those messages will temporarily affect your spam filing. But as soon as you mark them as spam your user Bayes filter will be fixed. Bill