Re: Does natural language skill translate to programming skill?

"Helmut" == Helmut Wollmersdorfer <helmut@xxxxxxxxxxxxxxxxx> writes:

Helmut> I switched from Prolog to Perl, because Perl is powerful
Helmut> and fast.

But Perl doesn't have a solver that automatically evaluates a predicate
given a set of rules and facts.



Helmut> My Perl scripts provide impressive results on text
Helmut> (written natural language), like information extraction,
Helmut> collect simple semantic nets, detect the use of synonyms
Helmut> and homonyms.

Perl is good and convenient for what it is supposed to be: Practical
Extraction and Report Language.

Prolog is good for what it is good for: perform deduction from a given
set of rules and set, employing _first order logic_.


Different tools, for different purposes. That fact that Perl suits
your specific needs better, doesn't mean that this tool is a better
tool for every purpose.


Helmut> But for parsing a complete sentence and do some reasoning
Helmut> with it, I would use Prolog.

Choosing the right tool is the most important decision, as the wrong
tool would make you fail, however hard you work.


>> I’m not aware of any natural language processing applications
>> written in Prolog or Lisp that have served a useful purpose
>> beyond a) absorbing research grant money or b) proving that
>> Prolog, Lisp and traditional computational linguistics
>> approaches are ill-suited to natural language processing.

Helmut> About 1987 I supervised a student, who did his diploma
Helmut> thesis on "application of conceptual graphs". His program
Helmut> written in Prolog was able to read in laws, and a user
Helmut> could enter an individual problem (case) related to this
Helmut> law, and the program solved the problem. The results were
Helmut> impressive. But it is easier to parse German texts than
Helmut> English ones.

Aidan was over-generalizing based on his ignorance. While Prolog is
not popular and generally unheard to the laymen, it does find its
application in many professional areas. Sometimes, the Prolog engine
is hidden so well behind the user interfaces (which can be text or
graphical) that the users aren't aware of it. (How many of you know
what a "p-n-p type transistor" is? Does that mean this is not found
in applications?)



Helmut> A friend of mine develops an artificial barkeeper. This is
Helmut> based on the idea of Eliza, and tries to do small
Helmut> talk. AFAIK it is written in Lisp.

Such programs have been in existence for decades. You'll usually find
one implemented in LISP (although, again, the user-interface or
Java/C++/C API 'skin' may have hidden the LISP 'guts' well inside).
LISP and Prolog are strong for "symbolic processing", as opposed to
"number crunching". That's why they're often chosen for AI programs.



--
Lee Sau Dan 李守敦 ~{@nJX6X~}

E-mail: danlee@xxxxxxxxxxxxxxxxxxxxxxxxxx
Home page: http://www.informatik.uni-freiburg.de/~danlee
.

Relevant Pages