4.1295 What is Computational Linguistics? (from LN) (1/127)

Elaine Brennan & Allen Renear (EDITORS@BROWNVM.BITNET)
Mon, 29 Apr 91 23:17:47 EDT

Humanist Discussion Group, Vol. 4, No. 1295. Monday, 29 Apr 1991.

Date: Sun, 28 Apr 91 22:44 EST
From: Jean Veronis <VERONIS@VASSAR>
Subject: What is Computational Linguistics?

I posted on LN the question recently posted on HUMANIST by the Crisp
Group ("What is Computational Linguistics?"), and here are the answers
(I didn't include Nancy Ide's, since it has been sent also directly to

Jean Veronis, LN co-editor

Date: Thu, 25 Apr 91 12:05:09 -0500
From: "Elizabeth A. Hinkelman" <eliz@tira.uchicago.edu>

Computational linguists are, roughly, linguists who use
computers and computer science, and computists who study
language. The computer becomes a laboratory for testing
models of language; a simple example is testing whether a
a set of grammar rules really generates the desired structures.

Viewing language analysis/generation as a computation with
well-defined computational properties enables discussion of
which properties enable it to account for the data. So
CL's have argued about whether morphophonology can be handled
with Regular Grammars (which don't reorder constituents and
so use only a fixed amount of memory, regardless of string length)
or require Context-Free grammars (can match constituents (like
nested parentheses), but need memory to store what's not matched
yet). The Regular Grammars can be used, with some inelegant
reduplication in the lexicon, for extracting word stems and
inflections in many languages.

For a non-linguist humanist, this means that it's possible to
build spelling checkers, concordance programs, and so on that
will recognize "doggies" as a diminutive plural of "dog".
Doing it exactly for a whole language requires tedious lexicon
building, but this is much easier than parsing English syntax
with a context-free grammar (a big CL occupation.) You can get
a notion of the state of the art from the Natural Language
Software Registry, available from registry@tira.uchicago.edu

Viewing language as computation also gives some intellectual
leverage on questions about interactive and processing
aspects of language. My area of discourse and pragmatics
has as a typical problem area question-answering; considering
how a computer could generate a helpful answer based on
some data base. This clearly involves information about
what intentions the asker has, and there are lots of
Artificial Intelligence people who would focus on this
aspect and not describe themselves as computational linguists.
Jerry Hobbs has a nice CSLI book on the implications of
this area for humanists.

Elizabeth Hinkelman,

Center for Information and Language Studies
University of Chicago

Date: Thu, 25 Apr 91 10:35:10 EDT
From: Michael Covington <MCOVINGT@UGA.CC.UGA.EDU>
Subject: RE: Q: What is computational linguistics?

I would define computational linguistics as the whole field of computer
processing or generation of natural-language utterances, provided the
goal of the processing is to deal with the meaning and/or linguistic
structure (including syntax, phonology, and the like). Purely statistical
studies of texts probably fall outside the field (unless some real
linguistic principles are being investigated by the statistics).

Much of computational linguistics falls within the realm of artificial
intelligence, but not all of it. Even if the artificial intelligence
enterprise were to be completely abandoned, there would still be
computational linguistics in a broader sense, for such things as the
computer testing of linguistic theories, imperfect natural-language
understanding, and the like.

- Michael A. Covington internet mcovingt@uga.cc.uga.edu -
- Artificial Intelligence Programs bitnet MCOVINGT@UGA -
- Graduate Studies Research Center phone 404 542-0359 -
- The University of Georgia fax 404 542-0349 -
- Athens, Georgia 30602 bix, mci mail MCOVINGTON -
- U.S.A. packet radio N4TMI@WB4BSG -

Date: Thu, 25 Apr 91 10:33:53 -0500
From: phall@ccwf.cc.utexas.edu

This is a reply to the question posed by the Crips Group at the
University of Arkansas. Very simply, Computational Linguistics (CL)
amounts to trying to make computers do linguistics. Of course, another
question then becomes, what is linguistics? There are several branches
to linguistics, and efforts have been underway to "computerize" all of
them. These branches--better, fields--are generally syntax, phonolgy,
phonetics, semantics, pragmatics, lexicology, and other areas (e.g.,
discourse analysis, linguistic-stylistics). CL has also helped
stimulate the area of mathematical linguistics and the area of logic.

I noticed that I got Crisp group spelled wrongly--sorry.

Hope that provides a simple explanation of CL.
Dave P-Hall

Date: Thu, 25 Apr 91 15:11:06 BST
From: Jock McNaught <jock@ccl.umist.ac.uk>

Un bon essai de definition de linguistique informatique, c'est l'article
(assez court) de Henry Thompson 'Natural language processing: a critical
analysis of the structure of the field, with some implications for
parsing', dans K. Sparck Jones & Y. Wilks (eds) (1983) Automatic Natural
Language Parsing. Ellis Horwood, publishers. pages 21-31.

Bien que cet article vise les problemes de parsing, la discussion de la
structure du domaine se poursuit au niveau general, et, a mon avis,
reussit a bien decrire les traits distinctifs de LI, et a distinguer LI
nettement de domaines et d'activites associes.


John McNaught                jock%ccl.umist.ac.uk@ean-relay.ac.uk (ean)
Centre for Computational     jock%ccl.umist.ac.uk@cunyvm.cuny.edu (arpa)
 Linguistics                 jock%ccl.umist.ac.uk@ac.uk (earn)
UMIST                        jock@cclsun.uucp
PO Box 88
Sackville Street
Manchester, UK               + (direct)
M60 1QD