Newsgroups: comp.ai.nat-lang
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!news.sei.cmu.edu!cis.ohio-state.edu!math.ohio-state.edu!cs.utexas.edu!swrinde!pipex!demon!news
From: Alexander Clark <asc@aclark.demon.co.uk>
Subject: Re: YACC / BNF Grammars for English
Message-ID: <Cw5quG.HrC@demon.co.uk>
X-Xxdate: Thu, 15 Sep 1994 07:26:36 GMT
Sender: news@demon.co.uk (Usenet Administration)
Nntp-Posting-Host: aclark.demon.co.uk
Organization: None
X-Newsreader: Nuntius Version 1.2
References: <Cw3DCt.Dw1@news.Hawaii.Edu>
Date: Thu, 15 Sep 1994 06:18:15 GMT
Lines: 25

In article <355ts6$l5b@hobbes.cc.uga.edu> Michael Covington, mcovingt@ai.uga.edu writes:
> : Obviously, English isn't directly Yacc parsable.  However, I remember seeing
> : somewhere that there are only so many noun/verb phrases and that Yacc can
> : do a fair amount of the parsing.  So, does anyone out there know
> : of some BNF / Yacc grammars which can handle English (in some form)?   
> 
> Not directly, but Gazdar, Klein, Pullum, & Sag's book _Generalized
> Phrase Structure Grammar_ (1985) will get you started.  They use something
> rather like a van Wijngarden grammar (rules generated by metarules)
> and what you will have to do is factor out the metarules and have a
> redundant PS grammar instead of a generalized one.

In COLING '94 there was a paper by Ishii, Ohta and SAito  
(An efficient parser generator for natural language)
Which describes an extension of Yacc called NLYacc that can deal with Tomita-style 
generalized LR grammars.  You can get it from nlyacc@nak.math.keio.ac.jp.
It comes (apparently) with source code and some grammars.


BTW with GPSG , isn't the derived CF grammar, though of course technically finite, 
very large indeed?
---------------------------
Alexander Clark
asc@aclark.demon.co.uk
---------------------------
