Machine translation interlinguae (was Re: Linguistic knowledge and conlanging (...))
|From:||Jörg Rhiemeier <joerg_rhiemeier@...>|
|Date:||Tuesday, July 27, 2004, 19:38|
On Mon, 26 Jul 2004 21:04:08 -0500,
"Mark P. Line" <mark@...> wrote:
> Jörg Rhiemeier said:
> >> > Or Old Albic would probably not be a good machine translation
> >> > interlingua, for example.
> >> No, you have to use Aymara for that.
> > Does this mean that Rick Morneau just wasted a lot of time and effort?
> On the contrary, Rick's approach has as good a chance of success for MT
> applications as any other I've seen. Even if it didn't, his work would
> still stand as a useful contribution to natlang semantics -- not a waste
> of time or effort by any stretch of the imagination.
I am not an machine translation expert, so I cannot judge the
adequateness (or inadequateness) of his language for what it has been
designed for. (Nor can I rate his opinion, laid out in an article
on Rick Harrison's site, that Lojban is NOT appropriate.)
And either way, Rick Morneau's monograph is impressive, and insightful
for conlangers. It teaches one not to take such notions as "subject"
and "object", or "active" and "passive" for granted, for instance,
and casts a light on a wealth of possible distinctions to make
in a conlang.
> I was being facetious about Aymara: somebody did actually propose its use
> as an MT interlingua. *shrug*
Of course. There is some kind of mythology surrounding Aymara:
it being the most logical of all natlangs, and all that.
What is really so special about Aymara that people tell such tales
about it? I have heard similar claims of it being the most logical
of all human languages about Latin. In a Stephen Baxter novel,
Latin is used to communicate with alien intelligences on the ground
of exactly that.
Well, Old Albic is hardly less "logical" than Latin:
its case marking is more precise (it is based on semantic
notions such as "agent" and "patient" and degrees of volition
- see my post on that subject - rather than formal notions such
as "subject" and "object"), its morphology more regular.
I can imagine people proposing it as an MT interlingua
in the world of the League of Lost Languages, where it is
a half-forgotten antique natlang. Nevertheless, "logicality"
(however defined) or appropriateness for machine translation
weren't design goals of it.