Theo Van Dinter wrote:
Well, we're off discussing this issue at the moment. We're going back and forth between giving out a working, but generic (ie: not terrific) set of data which you can use to create a database, and enabling autotraining (so that your incoming messages will train the system for you).
That seems reasonable. I'll just want to be able to feed files full of spam
to it, so system administrators have various ways to configure processing of this stuff. Personally, I'll probably make individual probably-spam and trash-spam folders available through IMAP and webmail for review and each user's trash-spam will be fed (automatically as well as on demand) to the bayes filter.
Incidentally, it seems to me the bayes filters are actually general identification filters. I haven't peeked at 2.5 so don't know if several bayes filters can be chained together for dropping mail in appropriate folders (corporate/support/maintenance/sales...). I specifically refer to "chaining" because I suspect they are implemented as pass/nopass filters rather than a multiple-category classifier where a single pass can identify to several categories.
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
