On Mon, 2 Apr 2001, Eric Scott wrote:
> Could you elaborate on this? I also find Langacker's
> writings very compelling; certainly his notions of
> 'entrenchment' of certain constructs suggest a statistical
> basis. It seems to me that he also provides a good bridge
> between corpus linguistics and symbolic/knowledge-based
> approaches. I would be interested in reading anyone who has
> done work in this area. Are there forums where this is/has
> been discussed? I would also like to know more about the
> nature(s) of the disagreements you mention.
If you want to see a corpus as just one of the many complex entities in
the world that we might want to work with in some way, then we can analyse
the corpus (and its annotations) using probabilistic models of various
kinds to describe patterns in the data. On the other hand, many linguists
are interested in the nature of the cognitive system that produces and
comprehends language and so then we must look for grammatical
representations that are congruent with the data patterns. (For the sake
of brevity I am assuming that the probabilistic models describe the data
and are not models of mental grammar.) I find Cognitive Grammar to be a
good framework for capturing the variability of corpus data even though it
is not corpus-based.
I will let others who disagree with the Cognitive Grammar approach say
their piece, but some people just don't like Langacker's diagrammatic
representations of grammatical info. Others may feel that corpus data
simply does not give us any evidence of the organisation of grammatical
information in the mind. And others may feel that notions such as
"schema" are useful in that we as linguists can handle them cognitively,
but that these sorts of representations are actually poor approximations
of network models.
If I can give a plug for a volume I edited with Suzanne Kemmer, you might
want to look at Usage-Based Models of Language. (It is in Borders
bookshops and so you can look at it while you have a coffee.) That volume
has a paper by Langacker and others such as Bybee, Lamb, MacWhinney and
Biber who give their slant on usage-based approaches.
You might also look at Schoenefeld (1999) Corpus Linguistics and
Cognitivism in IJCL 4, 1.
Michael
----------------------------------------------------------------------
Michael Barlow, Department of Linguistics, Rice University
barlow@rice.edu www.ruf.rice.edu/~barlow
Athelstan barlow@athel.com www.athel.com (U.S.) www.athelstan.com (UK)
This archive was generated by hypermail 2b29 : Wed Apr 04 2001 - 04:51:37 MET DST