Re: Proposal for declinations in gettext

[Yann Dirson <ydirson altern org>]

On Sat, Jun 14, 2003 at 05:43:05PM +0200, Martin v. Löwis wrote:
> The problem seems obvious to me: It is plain incorrect that "king" is
> a separate msgid, if it is meant to be pasted in different
> contexts. Instead, it should be added into any context where it is
> meant to be pasted into, forming separate msgids.
> 
> Telling apart declinations might be insufficient, because some
> languages may need changes to verbs, too, depending on whether it is
> "king" or "queen" that is being "move"d.

Yes, the problems mentionned are really just specific aspects of a
much more general case.  And since current tools are based on English
(which is a quite simple one) as central language, they may simply not
be sufficient.

I've already thought of this issue several times (about declinations
and plurals), and the approach that seems the most promising to me is
the one used by the UNL (Universal Networking Language)[1].

Basically, their project use a pivot language that is not english, but
the UNL they define.  The language is defined by a set of rules, and a
centralized dictionnary to which people are allowed to add vocabulary
items when needed.  Each of their national team produces a
fully-automated translator from UNL to their natural language, and a
translation-helper from their language to UNL (since the latter
process is known not to be possible to automate dur to context and
ambiguities).  In the past, some teams have provided public web access
to their UNL->natural-language (the spanish one at least used to be
available).

The main problem is that, although the UNL project is exists under the
aegis of the UN, they work in a somewhat closed way, and their
software is not available in source form, at least without being a
member of the organization.  You'll notice that many parts of their
website are password-protected.  Some of their central specs are
available, though, and we may chose to approach them with the idea of
making their stuff more universal by licencing it under a free
licence...  Then we could solve all problems in one move - if we can
convince programmers to use UNL strings in their programs...

[1] http://www.unl.ias.unu.edu/

-- 
Yann Dirson    <ydirson@altern.org> |    Why make M$-Bill richer & richer ?
Debian-related: <dirson@debian.org> |   Support Debian GNU/Linux:
Pro:    <yann.dirson@fr.alcove.com> |  Freedom, Power, Stability, Gratuity
     http://ydirson.free.fr/        | Check <http://www.debian.org/>