Re: [Hepserver-devel] Status of IMAP / integrating filtering

[Abe F. <ab...@fe...>]

On Mon, 2003-05-12 at 00:20, Gordon Weakliem wrote:
> Abe Fettig wrote:
> 
> >>2) I've been thinking about experimenting with integrating Bayesian 
> >>classification into Hep.  I've been studying the source for
> Spambayes, 
> >>but I'm wondering if/how it could fit into Hep's architecture.  I
> guess 
> >>the easiest answer is to put Spambayes' POP3 filter in front of the
> POP3 
> >>protocol - do you think that there's a way to get it all into the
> same 
> >>package?  I think that to get Bayesian filtering to work, you'd have
> to
> >>    
> >>
> >
> >My feeling is that it probably makes sense to put any kind of
> filtering
> >into the Hep core, rather than tying it to POP3 or another protocol. 
> >Could you give me a better idea of what you could use Bayesian
> filtering
> >for?  I know basically how it's used for spam detection, but to the
> best
> >of my knowledge nobody is using Hep to read e-mail at the moment...
> >  
> >
> The point is more managing information - Graham's paper was
> specifically 
> targeting spam but since then there's been various people talking
> about 
> extending the idea to just general classification.  My idea was 
> specifically that you could create categories and then have a program 
> tag incoming messages based on what category it thought they should be
> in.  I have no idea if it would work, but I've been kicking the idea 
> around and thought that it would be worth experimenting with.

So, for example, you could make a "Really Interesting News" folder, and
put messages that you thought were interesting in there, and then the
Bayesian filter would compare each incoming message to the contents of
that folder and determine if the message was Really Interesting or not.
Is that the sort of thing you'd want to do?