Re: controlled NLs
Rich,
There are many points that would require a
dissertation to address in detail, but I'll
just comment on a couple.
RC> ... bring some degree of realism to nlp
> question-answering systems, like simplistic
> agents in the appliances.
>
> Has anyone published work about using controlled
> nls in this kind of environment?
As I said to Jean-Luc, controlled NL and true NL
are totally different things that just happen to
look the same for some narrowly delineated subset.
Doug Skuce's ClearTalk, Norbert Fuchs's ACE,
Rolf Schwitter's PENG, and my CLCE are all designed
as more readable front-ends for knowledge engineers,
*not* for casual users.
RC> It occurs to me that a primitive-based controlled nl,
> using something like Wierzbicka's NSM primitives
> bound to specific synsets would be the kernel forms
> of a viable controlled nl.
No. Anna W. was most definitely not trying to define
a controlled NL for knowledge engineers, and she has
explicitly said that she has no intention of doing
anything with any computer-oriented logic of any kind.
Her primitives are very interesting from the point of
view of cognitive science, but not as a basis for a
controlled NL.
As far as teaching Cyc or anything like it, Lenat and
his group spent 20 years working on that problem and
talking with a lot of people who are have thought long
and hard about such issues. There are many alternatives
that might be tried, but I don't believe that any of
them will give a quick answer very soon.
Bottom line: Those people who are working on Controlled NLs
are taking a very tiny step that is easy to do with current
technology. Any such step is *insignificant* compared to
what is needed for true NL understanding. Please do not
blur the line between the two.
John Sowa