[Corpora-List] Chomsky and computational linguistics

[Rob Freeman]

I agree totally Steve. Can I just add a comment. Maybe you see this already,
but in case anyone is missing it. I want to emphasize that the solution may
be even closer than saying "I'll get Y and Z later, maybe when I have more
advanced techniques."

I think we can get whichever we want of X, Y, and Z, right now. The
essential missing insight is that we just have to understand we can't get
them all at once. Not all in one, internally consistent, grammar, anyway.

We can't get them all in one internally consistent grammar because X means
interpreting the data one way, and Y, and Z mean interpreting it in other
ways, ways which may actually contradict X when looked at globally (though
not locally): e.g. "black" is similar to "strong", but "black" is also _not_
similar to "strong".

I don't think the problem has been our techniques, which have been good for
50 years or more (until Chomsky noticed they gave multiple inconsistent
results.) The problem has been our goals. If we change our goal from that of
finding one complete grammar (and the functional, cognitive guys are just as
guilty of this, not to mention the engineers and machine learning guys), to
that of selecting (at will) between many possible inconsistent grammars, we
can solve syntax now (and reconcile it with semantics, function, usage...
etc.)

-Rob

On 8/1/07, Steve Finch <s.finch at daxtra.com> wrote:
>
> OK, now I can't resist....
>
> Mike Maxwell wrote:
> > Rob Freeman wrote:
> > > 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
> > > it.
> >
> > If they all fit the observations thus far, how would we choose among
> > them?  Unless you mean that I might control a number of dialects, and
> > can turn one or another on to make a point, or to make a joke.  (And in
> > fact, I might could.)  I doubt that most linguists (generative or
> > otherwise) would take issue with you on that.
>
> When we admit that there are multiple "structures" in language, each of
> which
> might have something interesting to say about language, the raft of
> arguments
> that Chomsky, Fodor et al originally proposed that essentially say that
> not
> *all* tranformational syntactic structure can be empirically inferred (woe
> betide theoretical physics, btw, for they have a yet harder problem
> without
> innate guidance) begins to crumble.  If we don't have to understand *all*
> of
> language in one go, then arguments that say "you might be able to get X
> that
> way, but what about Y and Z" can be countered by "I'll get Y and Z later,
> maybe when I have more advanced techniques that you or I haven't thought
> of
> yet, but right now I'm interested in X".
>
> And then we can say without controversy that there is a *lot* of structure
> in
> language that can evidently be - and has demonstrably been - found
> empirically.
>
> And in general I would say ask not what engineering can do for theory, but
> rather what theory can do for engineering.  The huge silence in this list
> on
> the latter point speaks volumes.
>
> - Steve.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/corpora/attachments/20070802/5ac507e3/attachment.htm>
-------------- next part --------------
_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora