|Index > Scribe > [new user] How do I file/organise email to enable Bayesian filtering?|
|Author/Date||[new user] How do I file/organise email to enable Bayesian filtering?|
|[long time i.scribe user, new inscribe user] :-)
I've looked at the documentation (inscribe 1.86/Win32) and the forums, but I don't seem to be able to see how I'm supposed to train the filter.
At the moment I've set up a "/Spam?" folder, and set the options to filter into this directory for training.
From reading other forum posts, I'm led to believe that confirmed spam is to be stored (and not deleted) in the "/Spam" folder.
So, my question is, how do I update the Bayesian corpus with the content of a downloaded mail so that when I activate the filter proper in live mode, it will do its thing? Or, more sussinctly, how do I filter spam?
Should I be dropping spam into the "/Spam?" folder to create the wordlist?
|Run the Filters -> Rebuild Bayesian Word Lists ?|
|So the steps are:
1. Drop known spam into the training folder (i.e. "/Spam?")
2. Select Filters > Rebuild Bayesian Word Lists
What about when the filter is live? Do I drop new spam into the "/Spam" folder? Can I safely move false positives out of this "/Spam" folder? Do I need to rebuild the wordlist after every mail move?
|I would use the "Delete As Spam" button on mail that are spam instead of drag and drop.
Then rebuild the word lists every so often.
If you find a false positive in the spam folder move it out into a normal folder (e.g. /Inbox) and rebuild the word lists.
|The training process involves 2 main things:
1) Delete as spam anything that IS spam.
2) Periodically scan the "Probably" folder and "Delete As Spam" any spam in there and move any non-spam back into your normal folders. Rebuild the word lists every so often. Gradually the mail arriving in the probably folder will become almost all spam at which point change the bayesian filter mode into "Live".