[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kmail-devel
Subject:    Re: Spam filter
From:       Ferdinand Gassauer <gassauer () kde ! org>
Date:       2002-12-16 19:57:22
[Download RAW message or body]

Marc Mutz wrote:

> On Monday 16 December 2002 19:40, Ferdinand Gassauer wrote:
>> Hi!
>> I just saw the Mozilla anouncement
>> http://www.mozilla.org/mailnews/spam.html
>> Is anything like that planed for kmail ?
>> or is it sufficient to set up the Bayesian filtering as a filter rule
>> (pipe) for kmail ? wihtout having a nice GUI?
> 
> Bayesian spam filtering is overestimated, like so many ANN techniques.
> 
> Because you are able to train it, you _have_ to train it. With example
> mails. _Lots_ of example mails. Hundreds or thousands.
> 
> And if the filter authors provide that initial training for you,
> Bayesian filtering is just like spamassassin, only that no-one
> understands the rules the Bayesian filter training comes up with.
> 
> That means that the biggest advantage of Bayesian filters - that they
> learn what is spam for _you_ and not the filter authors - is taken away
> from the system.
> 
> That said, we will probably still have a spamfilter interface in KMail
> 1.6 (KDE 3.2), due to popular demand. It also fits nicely with the
> Sieve'ification of KMail filtering, since Sieve has a spamtest command
> extension.
> 
> The current way to use spam filters is indeed to use "pipe through" and
> then check for the headers that the filter program inserted in a
> subsequent rule.
> 
> if <size> <is greater than> 0
> then pipe through "myspamapp --pipe-through-options"
> 
> if "myspamheader" <contains> "yes"
> then move to folder "junk"
> 
> Marc
> 
Hmm!
What's needed is a (configurable) "Move to spam" button which trains the
Baysian filter.
well I'll give spamassassin a try for now
-- 
cu
ferdinand
_______________________________________________
KMail Developers mailing list
kmail@mail.kde.org
http://mail.kde.org/mailman/listinfo/kmail
[prev in list] [next in list] [prev in thread] [next in thread]