Re: Universal Translation Language

Jim Henry wrote:
>
> My understanding of the "traditional approach" is:
>
> software translates source to interlanguage, then interlanguage to target
> language
>
> Yours seems to be:
>
> human translates source language to UTL, software translates UTL to target
> language
Well.  If I understand Marcos properly the sequence would be:
 1) Software converts surce into UTL
 2) Human make exhaustive revision of UTL (by far easier for a human than
UNL)
 3) Software translate from UTL to multiple target languages.
 4) Human revision of style.

If using UNL or other machine-but-no-human parseable language, steep 2)
should go after 3) once for each target language.  If the UTL->NL algorithm
is good enough and the UTL source is clear, the style revision is just that:
a style revision, and the human that performs it doesn't need the source (NL
or UTL).

There are then two problems to solve: create a good UTL which is human and
machine parceable, and create good NL->UTL and UTL->NL translators, this
last is suposed to be much easier than any NL->NL translator and the NL->UTL
doesn't need to be better than any of today's NL->NL translators.

-- Carlos Th