Computation...

Shalom Lappin lappin at dcs.kcl.ac.uk
Sun Sep 22 19:48:25 UTC 2002


Hi Bob,
   Thanks for this information on your implemented work , which I appreciate.
You will notice that I qualified my comment to Ash by saying that
most theoretical linguists working within the MP avoid specifying
the precise computational properties of the operations and models that they
posit. The comment was not intended to apply to computational linguists,
like you and Ed, who have tried to formalize and implement parts of the MP.
Interestingly, in Ed's papers on minimalist grammars (at least those which
I have seen either published or on his web site), he seems not to make
use of economy conditions of any kind. Do you follow a similar strategy?
In general, such conditions (global or local) have been fairly central
to the theoretical model of the MP, and so Ed's formalization seems to deal
only with the feature checking and (parts of) the bare phrase structure
components of Chomsky 1995. Regards.
                            Shalom
>
> Hi Friends,
>
> Well, there's  certainly plenty to diss minimalism about - and that I would
> agree with -
> but there are *some* people, like Ed Stabler and myself, that *have*
> worked out actual computations "in the sense of an explicit set of steps
> applying  to well-defined input and yielding a fully specified  output"...
>
> If you want to take a look at a parser, go to my student Sourabh Niyogi's
> web page here:
> http://web.mit.edu/~niyogi/www/minimal.htm
>
> And you can try it yourself -  it is free for download.
> you can download the Scheme code for yourself that will map strings to
> structured representations...
> What's more, we've extended this (straightforwardly) to include an EPP
> feature (thus accommodating scrambling - no apologies for
> the analysis, please - that's not what's at issue here), and a
> probabilistic parser as well
> (that fixes the feature values on the basis of input examples), using an EM
> algorithm...
>
>
> As for 'computation in the lexicon' -- here too we've tried some
> approaches; on that same page you can download Sourabh's paper (and ppt slides)
> describing how to fold in the Ken Hale incorporation story into this
> picture.  Whether
> that makes sense is another tale.  (Neither Sourabh nor I think so anymore
> - we'd
> rather put this information on the 'encyclopedic' side of the line...)
>
> So, with all due respect to my friend Shalom, 'bare of computation' is
> stretching
> it a bit (unless you want to exclude computationalists like Ed and myself,
> establishing the point by definition)
>
> Comments, as always, welcome.  We've actually gone far beyond this starting
> point in the past few months - if you are interested, take a look at
> Sourabh's paper
> on the Bayesian integration of syntax and semantics at
> http://web.mit.edu/~niyogi/www/
> But this would take us far afield from the main point here.
>
> Best regards,
> Bob Berwick
>
>



More information about the HPSG-L mailing list