Re: [Dspam-user] Training seems to be ineffective

[David R. <dr...@gm...>]

On Wed, Apr 10, 2013 at 12:49 PM, David Rees <dr...@gm...> wrote:
> I've been running dspam for a while and have a few users using it with good
> filtering rates, although I had noticed that on my account despite repeated
> training there are certain newsletters that continue to get tagged as false
> positives. I chalked it up to using TEFT mode for quite a while after some
> research and switched to TUM.
>
> But now I've started a new user which receives a good volume of extremely
> predictable SPAM and despite training these mails continue to be delivered
> with a typical Confidence of 0.9899 and Probability of 0.0000.

Following up on this list for the archives: The systems affected by
this were running PostgreSQL 9.2. But dspam was linked against either
8.4 or 9.0. Recompiling dspam against 9.2 seems to have solved the
problem - will do more testing to confirm for sure!

-Dave