Re: patch to add classifying threshold

On Thu, Apr 29, 2004 at 10:50:44AM +0200, xavier dutoit wrote:
> I did implement something roughly similar on my MDA, but outside of 
> ifile (I take the first two lines of the standard output result and do 
> the diff). Wasn't it possible to do that with procmail ?

yeah, like my first quick solution; but hacking the C src gots a faster 
and better solution.

> Do you use more than two categories (spam, ham) ? In that case how do 
> you deal this all the in-between categories (A, B, C, between A and B... 
> between C and B), as all the possible combinations becomes quickly 
> unmanagable, isn't it ?

that hack was somewhat talored to the spam,ham case, ie 2 cat.s; but ifile
sorts ratings and the patch always considers top 2 ratings, hence if you 
need to decide where to put a msg among several (+2) cat.s, usually you need
to know just the winning one, and maybe the next closest one. For more cat.s
I don't think you'd need to consider all combinations (well, depends on what 
you want to do), but I'd extend the patch to have ifile say eg:

categories: spam,ham,cheese,fruits...

ifile -T 20,3 -q -c msg

   spam
or
   spam,cheese
or
   ham,spam,fruits

ie '-T x,n': consider threshold x among up to top n cat.s.
kinda 'fuzzyfier'...


-- paolo
 
 GPG/PGP id:0x21426690 kfp:EDFB 0103 A8D8 4180 8AB5  D59E 9771 0F28 2142 6690