[Spambayes] progress on POP+VM+ZODB deployment

Jeremy Hylton jeremy@alum.mit.edu
Mon Oct 28 15:08:04 2002

>>>>> "DS" == Derek Simkowiak <dereks@itsite.com> writes:

  DS> In other words, Bayesian filtering (as popularized by the rticle
  DS> "A Plan for Spam") is only good for individuals, or small groups
  DS> of individuals who all like the same kinds of ham.

A single classifier is only good for individuals or for groups / lists
where there is a uniform notion of what is ham and what is spam.  The
general approach to filtering could certainly be used for a large
institution, but seems to require some tailoring to an individual's
ham.  There are a bunch of interesting UI and systems issues to
resolve for such usage.