Re: CG: Re: Controlled Natural Language vs. Natural Language
Rick and Paula,
All your points are valid, and I can't disagree
with any of them. I think that the main reason
why the lines are blurred is that NLs are so very
robust and flexible that they can accommodate an
extremely wide range of uses. Or in other words,
people are so very flexible that they can interpret
almost anything that uses their native vocabulary
in some rough approximation to their native syntax.
RW> In a sense, all NL interfaces with computers are
> Controlled Languages. They all rely on domain-
> specific word senses, and none of them allow the
> full range of English.
That's a good summary of what I was trying to say.
PN> One partitioning that might help is tri-partite,
> and is based ultimately on a test of the kind of
> response that might be obtained for a bad input
> to a language processor, as follows:
> (1) NFL: naturalized formal language...
> (2) ERNL: explicitly restricted NL...
> (3) IRNL: implicitly restricted NL...
That's a good distinction, and I think it would make
a good topic for a review article that analyzed a
variety of systems to show how they relate to NLs
and to the underlying methods of processing.
But as you say,
PN> I might note that the language of a text might
> fall into more than one category, depending on the
> kind of processor to which it is submitted....
There will inevitably be systems that cross the lines,
especially given Peter Clark's suggestion that NFLs,
as defined above, could be made more helpful and
forgiving by relaxing the constraints. That would
move them closer to the IRNLs in category 3.
I'd interpret these issues in terms of Wittgenstein's
language games: people are extremely flexible in the
range of language games they play among their peers
and among "outsiders", such as children, foreigners,
and even computers.
John