Ken,
I erased the DSPAM database, reconfigured it to toe/osb, and retrained. Now
it's working much better. Spam hit rate is >90%.
Thanks!
>
--
___
Dspam-user mailing list
Dspa
On Tue, Sep 06, 2016 at 10:40:15AM -0700, Jerry Gardner wrote:
> On Tue, Sep 6, 2016 at 10:30 AM, Kenneth Marshall wrote:
>
> >
> > Yes, just delete them. You can look at /etc/dspam.conf to see where they
> > are stored.
> >
>
> Okay, thanks. I've saved all spam and non-spam I've received since
On Tue, Sep 6, 2016 at 10:30 AM, Kenneth Marshall wrote:
>
> Yes, just delete them. You can look at /etc/dspam.conf to see where they
> are stored.
>
Okay, thanks. I've saved all spam and non-spam I've received since I set
this server up. Should I use it to train DSPAM before I change teft to to
On Tue, Sep 06, 2016 at 10:06:11AM -0700, Jerry Gardner wrote:
> On Fri, Sep 2, 2016 at 1:14 PM, Kenneth Marshall wrote:
>
> > Hi Jerry,
> >
> > First, I see that you have teft set as the training mode. Change that
> > to toe, instead. Otherwise your accuracy can degrade over time. Second,
> > yo
On Fri, Sep 2, 2016 at 1:14 PM, Kenneth Marshall wrote:
> Hi Jerry,
>
> First, I see that you have teft set as the training mode. Change that
> to toe, instead. Otherwise your accuracy can degrade over time. Second,
> you are using the hash driver and the chain tokenizer, even though the
> commen
Yes.
On Fri, Sep 2, 2016 at 1:14 PM, David A. Desrosiers <
david.a.desrosi...@gmail.com> wrote:
> On 9/2/16 3:44 PM, Jerry Gardner wrote:
> > What can I do to help DSPAM learn faster so I can cut down all of the
> > spam I have to manually deal with each day?
>
> Did you feed it the initial corpu
Hi Jerry,
First, I see that you have teft set as the training mode. Change that
to toe, instead. Otherwise your accuracy can degrade over time. Second,
you are using the hash driver and the chain tokenizer, even though the
comments in the config file recommends osb instead. Make those two
changes
On 9/2/16 3:44 PM, Jerry Gardner wrote:
> What can I do to help DSPAM learn faster so I can cut down all of the
> spam I have to manually deal with each day?
Did you feed it the initial corpus to train it with both ham and spam?
I set up a new mail server about four months ago with DSPAM as the spam
filter. While it is filtering out some spam, it completely misses most of
it. I trained it for the first 2500 emails, but it doesn't seem to be
getting any better at filtering after that. I regularly retrain any false
negatives
