From: Ka-Ping Yee (ping@lfw.org)
Date: Sun Apr 23 2000 - 21:52:37 PDT
On Sun, 23 Apr 2000, Dan Kohn wrote:
> I thought I learned in CS 101 that creating a universal language wasn't
> possible, because (for example) to describe the difference between two words
> in Yiddish could take a couple pages of English text.
This is an interesting-looking project. Many of the machine-
translation projects i've heard of involve some kind of central
semantic representation of the language -- which seems like the
coolest part to me but also the most intractably difficult. I
would consider universal semantic representation the "Holy Grail
of machine translation". I suspect you are right that this
semantic representation is very very hard to derive from natural
language.
That is, i think what they call "EnConversion" (natural language
to UNL conversion) is arbitrarily hard.
However, the idea of *starting* from a central representation
like UNL is fairly interesting. I haven't thought about just
separating that side of the issue before. But it seems more
reasonable to me that you could invent a language like UNL that
is carefully designed to be limited to simpler and easily
composable concepts -- one which doesn't have the expressive
power and aesthetics of a natural language, but is fairly
easily convertible into most natural languages. And then, one
could presumably learn to compose text in UNL, and you could
write a single document that was adequate for automatic conversion
to many languages (not ideal for any one language, but adequate
for most).
And then countries could write treaties in this language,
and ACM programming contest or International Math Olympiad
problems could be composed in this language, etc. :)
I don't know. Have i just described Esperanto or Loglan?
-- ?!ng
"The only `intuitive' interface is the nipple. After that, it's all learned."
-- Bruce Ediger, on user interfaces
This archive was generated by hypermail 2b29 : Sun Apr 23 2000 - 21:55:06 PDT