Re: Chomsky and computational linguistics
2007-08-01 01:46:05 GMT
On 8/1/07, Mike Maxwell <maxwell <at> umiacs.umd.edu> wrote:
Rob Freeman wrote:
> Now, to Chomsky that meant something in the mind of a native speaker
> must select between them.
> I don't think that hypothesis panned out.
This is the "50 years is a long time" thing.
Or course, that one hit innate solution may be around the corner. But we've turned a lot of corners.
There is also now quite a large body of work bearing out just how important all the detail is. Not least corpus linguistics, but also Functional and Cognitive linguistics. They all started from different positions, and there has been a general movement to the importance of usage and detail.
When you think about it Functional and Cognitive linguistics started out as far apart as you can imagine. Functionalism was close to behaviorism and denied mental reality. Cognitivism was just a branch of Universal Grammar and probably the most extreme statement of the anti-behaviorist position. Now we have some consensus, and the consensus is that detail matters.
And there is also the fact that if you explain the data another way, you simply don't need to posit an innate selection mechanism any more.
> However there is another way to interpret the same data. Same data,
> different conclusion.
> To me the fact we get many grammars from the same set of observations
> (observational insufficiency) means they are all good, and we need
> to keep the observations so we can find the one we need, when we need
If they all fit the observations thus far, how would we choose among
It is not a problem in practice. Take the one which best answers the question you want answered at any given moment. Is "black" in the same class as "strong"? Check if your context is "coffee" or "cloud".
It is only a problem when we try to squash all the grammars into one. When we do that we get something very much like the randomness of a statistical grammar, because sometimes "black" is the same as "strong", and sometimes it isn't.
_______________________________________________ Corpora mailing list Corpora <at> uib.no http://mailman.uib.no/listinfo/corpora