Re: dspam question

[Robert <http://dummy.us.eu.org/robert>]

--- http://www.outformations.com/~dek wrote:
> Hey Robert,
>    I read the below in the faq.   I think this is what is happening to me. 
>  I don't think I can do the mbox thing easily, what about the other 
> options?  Any advice?

I read this a while ago, too.

By "the mbox thing", do you mean feeding in a corpus?  I wish I could donate
some "innocent" mail to you, but don't think that my innocent mail is the same
as your innocent mail :-).

Have you tried deleting all your data and started afresh?  I forgot whether we
tried that or not.

My dspam is working surprisingly well.  It actually caught one spam that
"slipped through" my other filters.

Here are my dspam_stats:

robert            TS:     1 TI:    32 SM:     0 IM:     0 SC:   219 IC:  1760

TS = 1 -- that's my one spam.  Note that the way my system works is that I feed
in my spams as a corpus weekly.  These are the spams that my filter marked as
"spam".  I also feed in innocent mail as corpora when (1) whenever I send mail,
(2) the ham that I receive, (3) any Usenet messages which I read/download, and
(4) all the web pages that I regularly monitor.  (That's why the number is so
high.)  (BTW, dspam isn't the only anti-spam filter that I train.  I also train
ifile, bogofilter, and crm114 in addition.)

You know, one thing you might try is, using SQL magic of some sort, try copying
all of David's data to your own.  He'd be a good subject 'cause the e-mail he
receives is probably similar to the e-mail you receive.

> Cheers,
> Dale
> 
> Q. I've fed DSPAM thousands of spams, and am only getting marginal 
> accuracy. What's up?
> A. Your problem might be that you've fed DSPAM thousands of spams, but