Re: DO NOT REPLAY (was: REPLY ASAP.)
On Tue, May 20, 2003 at 05:17:06PM +0100, Pigeon wrote:
> On Tue, May 20, 2003 at 11:21:30AM +0200, David Fokkema wrote:
> > On Tue, May 20, 2003 at 10:58:17AM +0200, Nicolas Kratz wrote:
> > > On Tue, May 20, 2003 at 10:31:55AM +0200, David Fokkema wrote:
> > > > As baloo suggested this weekend: has my spamassassin autolearned this
> > > > thing as ham by (doubly) quoting it in full?
> > >
> > > Mine has not. It got a score of -0.2, which does not trigger ham
> > > autolearning. Spamassassin adds "autolearn=[spam|ham]" to the
> > > X-Spam-Status header if autolearn was triggered, if you don't see
> > > anything there, nothing happened.
> >
> > Ah, I see, :-(. Mine got scores of appr. -14. And yes, autolearn=ham.
> > Great. So I throw this in a spam folder and let spamassassin manually
> > autolearn this as spam? Guess so, <sigh>. I know now what to look for,
> > thanks!
>
> I wonder how people's Bayesian filters / autolearners have responded
> to the subjectless thread from Andrey Yashyn. I hit next-unread in
> mutt's pager and as soon as the text came onto the screen my mental
> spam filter gave it a large positive score. I thought it was another
> Nigerian-type scam, in fact. I was just about to fire it off to
> spamcop when I noticed that it was (a) short and (b) had a real
> request for Debian hidden in it. I wonder how many automatic systems
> gave false positives?
I thought exactly the same! I didn't even know why I bothered to read it
to the end. Spamassassin gave it a score of 2.6, at least here, and 5.0
is needed for a positive. So, spamassassin basically did a better job
than me, :-)
David
Reply to: