Newsgroups: comp.ai.nat-lang
Path: cantaloupe.srv.cs.cmu.edu!das-news.harvard.edu!news2.near.net!MathWorks.Com!europa.eng.gtefsd.com!howland.reston.ans.net!cs.utexas.edu!uunet!bcstec!bronte!snake!rwojcik
From: rwojcik@atc.boeing.com (Richard Wojcik)
Subject: Re: Books on Intro. natural language proces
Message-ID: <1994Sep21.155616.22367@grace.rt.cs.boeing.com>
Sender: usenet@grace.rt.cs.boeing.com (For news)
Reply-To: rwojcik@atc.boeing.com
Organization: Research & Technology
References: <35k3j5$43h@redwood.cs.scarolina.edu>
Date: Wed, 21 Sep 1994 15:56:16 GMT
Lines: 44

In article 43h@redwood.cs.scarolina.edu, rolandi@andy.hssc.scarolina.edu (Walter Rolandi) writes:
>flanhart@diku.dk (Rene Andersen) writes:
>
>>I would strongly NOT recommend Grishman. I think it's outdated, boring
>>and completely without visions.  To Grishman NLP means Natural
>>Language Parsing.  It's fundamental view - that language is a prurely
>>formal (syntactic) phenomenon - doesn't bring us any further. But, off
>>course, it can be read as an intruction to parsing. But I don't think
>>it is fruitfull to consider it a book to give a proper account of the
>>state of art.
>
>what is the state of the art now?  when i first became interested in
>nlp, the majority view was that of the symbol pushers.  
>
>has there been any significant movement toward the realization that
>nlp is actually a problem of social modelling which is completely
>dependent on an underlying adaptive learning ability?

Unfortunately, it is very difficult to get computers to do anything other than
push symbols.  I have had long conversations with my computer.  Cursing, cajoling--
nothing seems to work.  So I fall back on symbol pushing.   The machine, at
least, is happy with a more superficial approach to language.  ;-) 

Seriously, though, we need to pay more attention to the original question.  What
kind of information ought to go into an introductory text, as opposed to a 
treatise on an advanced subject?  I think that a substantial portion of that
text needs to address parsers and grammars.   That is what one expects NLP
experts to know a lot about.  That is what you need to know in order to pursue
a career in NLP.  It also helps you to understand the inadequacies in popular
techniques if you know some details about those techniques. 

The points you raise about "social" (user?) modelling and learning are very
good ones.  NLP applications, such as they are, sometimes do attempt to adapt
to different needs of different users.  And it would certainly be far cheaper and
more convenient to build programs that learned language.  We just don't know
how to  do that very well yet.  So why spend a lot of time on the advanced stuff
in an introductory textbook?

---

Disclaimer:  Opinions expressed above are not those of my employer.

    Rick Wojcik   (rick.wojcik@boeing.com)   Seattle, WA

