Re: CG: Fw: Intro to natural language processing
John F. Sowa wrote
> Rich,
>
> I checked Google for the phrase "Rob Freeman" and
> it led to the "Chaotic Language" web site:
>
> http://www.chaoticlanguage.com/
> Chaotic language - home
>
> My short take is (1) his comments about NLP being
> difficult are not wrong, but (2) I wouldn't hold
> my breath waiting for his proposed solution.
>
> John
Yes, thanks for the URL; he clearly has less to offer
than I would like, but I posted his message because
he summarized very well what I've been hearing from
so many people - a pessimistic view of the practicality
of nlp applications.
But that view reminds me of the many predictions
you've heard just before a market takes off big:
-Western Union thought the telephone would never
challenge the telegraph;
-Eventually half of the population will be employed
as telephone operators;
-There is a market for perhaps three computers in the US;
and many more of the same.
It seems to me that progress in NLP apps has been very
slow, but that we are getting into the fast accellarating
part of the S curve now. Computers are large, fast, and
getting more so every year. The amount of English
text on the web is huge, though not well organized.
Linguists are analyzing many niche areas of the technology,
such as discourse representation theory, the CoreLex
thesis, etc.
But pessimistic views like Freeman's remind me that
the amount of work needed to implement a marketable
system is still very, very large. I don't want to get into
a decades-long project before there's a payoff. Most
projects have to break even within two to three years to be
reasonably considered by investors, and also show potential
for fast growth and market dominance a few years after
that.
A few thoughts,
Rich