"big hungry joe" <········@spamhole.com> writes:
> I have heard that LISP is really good to know if you are interested in NLP.
> Cna someone tell me why?
The short answer is that most other languages don't have a primitive symbol
type, while Lisp does. And natural language is about symbols, not numbers.
You should try a short experiment. Write a toy NLP application in Lisp and
compare it to one in C++ or Java, then see what you think.
>>>>"KMP" == Kent M Pitman schrieb am Sun, 25 Mar 2001 06:16:05 GMT:
>> I have heard that LISP is really good to know if you are
>> interested in NLP. Cna someone tell me why?
KMP> The short answer is that most other languages don't have a
KMP> primitive symbol type, while Lisp does. And natural language is
KMP> about symbols, not numbers.
Yes, that is indeed one reason. It is however not necessary to
restrict oneself to Lisp, other "symbolic" languages might be useful
as well. Prolog, for instance, comes to mind.
In our department we use a two-fold approach (don't ask why, the
decision was made before I came here); the syntactic parser is
implemented in Smalltalk (actually using an actor language extension)
the knowledge representation is done in Loom with a mix of Loom and
Lisp functions for the semantic interface.
On a related topic, I found the thread about parallel lisp
applications very interesting; as one might expect, large scale NLP is
quite expensive. I think the Smalltalk part of our system might be
quite easy to port to a multiprocessor environment (at least the
parsing uses an asynchronous message passing protocol), but when it
comes to the Lisp part, we're pretty much stuck, I think. Has anybody
done work on a parallel implementation of a KR-system such as Loom (in
Lisp, of course)?
Holger
--
--- http://www.coling.uni-freiburg.de/~schauer/ ---
"Perl was the first language to integrate the
Unix environment into itself."
-- Per Abrahamsen in comp.emacs.xemacs
Holger Schauer <··············@gmx.de> writes:
> >>>>"KMP" == Kent M Pitman schrieb am Sun, 25 Mar 2001 06:16:05 GMT:
>
> >> I have heard that LISP is really good to know if you are
> >> interested in NLP. Cna someone tell me why?
>
> KMP> The short answer is that most other languages don't have a
> KMP> primitive symbol type, while Lisp does. And natural language is
> KMP> about symbols, not numbers.
>
> Yes, that is indeed one reason. It is however not necessary to
> restrict oneself to Lisp, other "symbolic" languages might be useful
> as well. Prolog, for instance, comes to mind.
big hugnry joe did not ask whether other languages were also good, so I
did not answer on that basis.
I did not suggest that prolog was bad for natural language processing,
in part because I have no really considered belief on the subject at all,
but partly also because my intuition would be that it would be ok--to the
extent prolog is ever good for anything, which we could debate, but please
let's not. This isn't the forum for it.
On Sat, 24 Mar 2001 18:27:15 -0600, "big hungry joe"
<········@spamhole.com> wrote:
> I have heard that LISP is really good to know if you are interested in NLP.
Here is a NLP system written in Lisp:
Grammar Writer's Workbench for Lexical Functional Grammar
http://www.parc.xerox.com/istl/groups/nltt/medley/
Paolo
--
EncyCMUCLopedia * Extensive collection of CMU Common Lisp documentation
http://cvs2.cons.org:8000/cmucl/doc/EncyCMUCLopedia/
There's also SNePS (Semantic NEtwork Processing System) written in
Common Lisp. The web site for it is
http://www.cse.buffalo.edu/sneps
--
Fred Gilham ······@csl.sri.com
[My tutors] got bored sooner than I, and laid down a general rule
that all statements about languages had to be in a higher level
language. I thereupon asked in what level of language that rule was
formulated. I got a very bad report. -- J. R. Lucas