Corpora: Chomsky and corpus linguistics

[Amanda Schiffrin]

I too have been watching this thread with interest - I wouldn't
like to have to answer any of the arguments point by point (I'm
sure my background knowledge wouldn't be up to this anyway!) -
but I would like to add my perspective to the Corpora melting
pot (and I'd be very interested in people's comments).

I've always seen the Generative vs. Corpus Linguistics debate
as analogous to Top Down vs. Bottom Up approaches (in Computing
for instance), or Cognitive Modelling vs. Behaviourism (in
Psychology).  Surely one is just as valid as the other?
(There's nothing like living on the edge, making statements
like that around here... ;-)

I've always thought that generative theories which do not
match or cannot be backed up by evidence in 'real' conversation,
should be discarded without a second thought (always assuming
such theories have been developed with the purpose of explaining
phenomena in human discourse in mind).

That is not to reject the value of developing such theories -
that's not what I mean at all.  For if you were to find a
cognitive model that does not contradict data to be found in
speech corpora, I believe you would be taking a substantial step
towards understanding the underlying structures and mechanisms
of language itself.

Far from seeing these two approaches as totally disconnected
fields of study, I see generative and corpus linguistics as
two approaches to the same problem.  There must surely be a
point at which both 'sides' will eventually meet?

I count myself as neither from one discipline nor the other;
neither empiricist nor pure theoretician, but some sort of a
hybrid.  I reason that one should inevitably complement the
other.

Mandy

-------------------------------------------------------
 Amanda Schiffrin         |
 AI Laboratory            | Tel: +44 (0)113 233 6818
 School of Computing      | Fax: +44 (0)113 233 5468
 The University of Leeds  | www.comp.leeds.ac.uk/mandy
 LEEDS    LS2 9JT    UK   |
-------------------------------------------------------