[Debtags-devel] Using Python NLTK for tag generation [was: AI for tag generation].

[Gustavo Franco]

Hi list,

I've joined debtags group at alioth and after doing some tagging i'm
now researching about how we can improve the current system.

Let me say what i've in mind:

- python-debtags : I'm planning to write python bindings for
libdebtags and i guess that it will be impossible to do using pyrex
but i can try with swig or something more complicated.It's needed to
make a better work with nltk (see below) and can be used by other
codes, of course.

- Python Natural Language Toolkit[0] : The bayesian idea sounds great
but we can use a library like ntlk to do the classification easily and
you can see that looking the documentation. Ok, we won't start with a
spam filter and just needing write some hacks here and there but i
think that nltk will be more flexible and useful for us.

By the way, python-debtags when started will be maintained on arch
(revision control).I'll keep it in house for some time and after that
i'll request a repository at arch.d.o.Comments?

I want to start writing python-debtags this weekend, and maybe
packaging python-nltk too to do some tests.I want to hear suggestions
and critics before. :)

[0] = http://nltk.sourceforge.net/

Thanks,
Gustavo Franco -- <stratus@acm.org>