Tom Emerson wrote:
[...] I'm not sure how to get SA to "learn" from what she classifies as ham/spam. She is using outlook express, so getting pristine copies
of message to pass to sa-learn will be difficult at best. [further munged by the fact that SA is running on my server for which she essentially doesn't have shell access...] Any thoughts on how to easily train messages with this setup?
Could she use IMAP to access spam training folders on your server? If so, she could simply drag spam/ham to the appropriate folders... maybe only while online. This is the easiest variation I've found.
If IMAP is not an option, create folders for ham and spam on her desktop and teach her to drag messages to the appropriate folder for classification. Creat a clickable batch file or something similar (unison? ftp?) to get those up to your server when she's online.
Finally, this is family, so normal rules don't apply. If she's OK with
it, have any "borderline" messages go to a folder you monitor and
classify. Have blatant spam go to another for automatic training. I did
this with my kids, and it's worked well. I selected a few
"authoritative" tests that automatically drop messages into spam
training folders (pyzor, razor, bayes_99) where they get trained then zapped. Others (anything scored by SA 5-12) I review for classification. Other tests result in a copy being filed for review. It's unobtrusive to them, yet easy to maintain for me.
I'd avoid any auto-learning as she's very likely to mis-classify many due to the hassle at times.
Just a few thoughts for a "friends and family" approach.
- Bob
