Re: Improving classification of spams

On Mon, 13 Jan 2003, Jason Rennie wrote:

>
> [email protected] said:
> > Booker Bense <[email protected]>:
> >
> > > - So far, I've had very good luck with the spam/non-spam issue,
> > > however I have hybrid system where I index everything, but only
> > > use ifile as a last resort. (i.e. I have a bunch of prefilters
> > > based on the sender/headers if none of those match see what ifile
> > > suggests. ) Every message is indexed by ifile after it gets
> > > filtered.
> >
> > i discontinued doing this thinking that i'd prevent ifile to see some
> > of the spam and therefore not training it correctly.
>
> Using ifile the way Booker does is perfectly fine.  It's actually better
> for the overall system performance if you first remove mailing lists and
> e-mails that you can filter deterministically.  The key is to ensure that
> you don't train ifile on such e-mails.  There's no chance that you'll be
> using ifile to filter mail, so allowing ifile to filter there will only
> increase the number of errors.
>

- I actually have ifile index the email lists. So far, it's been
more useful than painful[1] and since there are some lists that
get spam I an intending to use ifile on those lists as well,
haven't written the code yet though.

- Booker C. Bense


[1]- I have a few low volume lists that I aggregate into a single
mail folder, ifile works great for doing this automagically. I
find that I'm having to do a lot less tweaking of my filters to
compensate for wierd mailling list software. Right now, when I
join a list, I wait and see what ifile does with it. If it does
the "right" thing, then I don't bother with adding new filters.