Jerome Delamarche wrote:
Hi,
I'm configuring SA and I'm looking for an easy way for the end users to
improve their own Bayesian filters.
Users do not have interactive account on the Linux servers. They cannot use
"sa-learn" or any other Linux tools.
It could be fine if they could automatically resend to their own mailbox
spams not been filtered by SA.
SA could (?) determine it has already analyzed the message and automatically
consider it was a previous spam.
Then it could use the "auto-learn" feature to add it to the user spam
database ?
Or is there another way to do it ?
If your users can use IMAP, you can create a special folder where they
copy spam messages. The Linux server can sa-learn from these folders.
Or, you can use a system on the Linux server, such as Maia Mailguard,
that temporarily stores all message on the server and provides a
web-interface for user training.
Another option is to provide a special address that users forward spam
messages to. The main problem here is that the message must be
forwarded as an attachment in a way that a script on the Linux server
can extract the attachment and get something reasonably close to the
original spam. Thunderbird does a pretty good job with this. Outlook,
not so much.
-Stuart
