Re: Universal Translation Language
From: | Charles <catty@...> |
Date: | Sunday, May 30, 1999, 19:36 |
Marcos Franco wrote:
>
> On Sun, 30 May 1999 07:47:44 +0100, "Raymond A. Brown"
> <raybrown@...> skribis:
>
> >>I thought it was needless to say that UTL language is not intended for
> >>translation of literature, artistic texts, etc.
> >
> >Why?
> >
> >What then is the not-so-universal TL language intended for. Translating
> >scientific and/or technical texts? But such texts are not noted for having
> >a great deal of ambiguity?
> >
> >What is the "UTL" intended for?
>
> Ok, I'll try to explain it as clearly as I can.
>
> A lot of books, magazines, web pages, articles and documents of any
> kind are written every year. Most of them will not ever be translated
> to other languages, because of human translation costs. Translation
> software has been tried as a mean to reduce translator's work (and
> cost), but its output is usually too erroneous to be of help.
>
> However, with UTL, things can change substantially, as UTL is a
> language which can be processed easily and almost error-freely by a
> machine translator. As results of this, from a UTL version of any text
> one will be able to obtain several MT-ed versions of it to several
> languages with almost no need of human post-editing. Thus, translation
> costs are normally reduced to a 5-20% (depending on number of target
> languages and requirements of quality, given by post-editing).
I thought it was even simpler: UTL would be trivially easy to
translate *to* any natlang. This, because of UTL's regularity
and lack of accidental ambiguity. In any language, even UTL,
one can always deliberately ambiguize by saying "the sorta-XXX
kinda-YYY thing", so it isn't a thought-prison ...
UTL could be either written directly as the source text
pre-translation, or it could be roughly translated from
the writer's natlang and then proofed/corrected until the
natlang->UTL->natlang output looked correct to the writer himself;
then published simultaneously for all supported langs. This is
what the funded "UNL" project claims to be doing, I think.