We currently use DSPAM to front our trouble ticket system and
have found that a statistical filter is way, way less work and
less prone to false positives than ad hoc rule systems. It also
supports our favorite backend database. It sounds good to just
say ""Casino slots" in the title is a pretty easy catch..." but
when you start having thousands of rules, it gets unmanageable
pretty quickly. My two cents on rule-based versus statistical
filtering.
Cheers,
Ken
On Fri, Feb 22, 2008 at 09:25:33AM -0500, Andrew Sullivan wrote:
> On Thu, Feb 21, 2008 at 09:26:18PM -0400, Marc G. Fournier wrote:
> >
> > Are there any X-Spam headers being added to the messages? Are those
> > messsages scoring low?
>
> Yes and yes. I just got two more (casino ones, in this case) with spam
> scores similar to what others have posted.
>
> I'm wondering whether something simple and rule-based might help catch a lot
> of this before we start Bayes work on them. I mean, "Casino slots" in the
> title is a pretty easy catch :)
>
> A
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 6: explain analyze is your friend
>