[Debtags-devel] AI Tagger

Gustavo Franco gustavorfranco at gmail.com
Mon Aug 15 21:15:57 UTC 2005

Hi list,

Can't we start using results produced by AI tagger and bayesian as
'proof of concept' of multiple vocabularies ? My proposal is simple:

- Run your code(s) against the unstable packages trying to tag all
them or a subset, but document it anyway;

- Provide a interface (web/mail) to manual tag editing and recruit
your code(s) to not ignore this manual edited tags in a second run;

- Provide a vocabulary repository (Install debtags packages and check

- Let us see how we can improve each vocabulary and what's the best
(manual only, AI tagger + manual or bayesian + manual).

If you think that it can be done this way i'll add that a d-d-a should
be done, to let the DDs that weren't at debconf 5 know more about
debtags project, his status on Packages and finally the 'multiple

Gustavo Franco -- <stratus at debian.org>

More information about the Debtags-devel mailing list