Newsgroups: comp.ai.nat-lang
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!news.nic.surfnet.nl!sun4nl!freya.let.rug.nl!let.rug.nl!vannoord
From: vannoord@let.rug.nl (Gertjan van Noord)
Subject: Re: grammatical repairs and context-freeness
Sender: news@let.rug.nl (News system at let.rug.nl)
Message-ID: <1995May17.082731.14238@let.rug.nl>
Date: Wed, 17 May 1995 08:27:31 GMT
Distribution: na
References:  <DOWDING.95May16191013@Gansett.ai.sri.com>
Nntp-Posting-Host: saga.let.rug.nl
Organization: Faculty of Arts, University of Groningen
Lines: 40

In article <DOWDING.95May16191013@Gansett.ai.sri.com>, dowding@ai.sri.com (John Dowding) writes:
 > 
> 
> I've been looking at grammatical repairs for a while, where some 
> sequence of words in the utterance is repeated to correct some
> mistaken word:
> 
> <who> who needs to go to the mall.
> <who wants> who needs to go to the mall.
> <who wants that> who wants to go to the mall.
> <who wants to take> who wants to go to the mall.
> <who wants to go for> who wants to go to the mall.
> 
> This sure looks like the copy language to me, and the
> length of the copied region is unbounded.  

that's not clear to me. How do you show that it's unbounded?
Especially since it seems related to performance..

Furhtermore, lots of examples can also be found in which you don't
have to copy the previous words. So, maybe this copy language is
just a subset of essentially a finite-state mechanism which simply
says that you can restart anywhere you want... 


> 
> How persuasive is this argument for the non-context-freeness of
> english?  I realize this raises lots of issues about performance
> vs. competence.   While these sentences do contain performance errors,
> the speakers are using very rule-driven correction strategies.

cf. above. To be persuasive it is not just enough to point at a subset
of English that looks context-sensitive, but you have to give some kind
of intersection argument, I guess...

> Certainly the knowledge of these correction strategies must be
> considered part of the speaker's competence.
> 
agreed

