Query: Looking for natural language grammar

markjan@cs.kun.nl (Mark-Jan Nederhof)
Wed, 26 Aug 1992 11:41:31 GMT

          From comp.compilers

Related articles
Query: Looking for natural language grammar markjan@cs.kun.nl (1992-08-26)
| List of all articles for this month |

Newsgroups: comp.compilers
From: markjan@cs.kun.nl (Mark-Jan Nederhof)
Organization: University of Nijmegen, The Netherlands
Date: Wed, 26 Aug 1992 11:41:31 GMT
Keywords: LR(1), parse, question, comment

For our research, we are looking for a large grammar for natural language;
this grammar should be a context-free grammar possibly extended with
parameters and features of any form. The parameters and features are not
important to us, nor is the language generated by the grammar or the
declarative "meaning".

Our purpose is to measure the size of different (variants of) LR parsers
for natural language grammars. In particular, the influence of epsilon
rules (which sometimes make normal LR parsers or Tomita parsers loop) on
the size of LR tables is of great interest to us. The ``nastier'' the
grammar is for computational reasons the better it is for us. (We would
prefer a grammar which has been written without any regard for its
computational properties.)

We are looking for an appropriate grammar for a long time now. If you
have one, then we would be extremely grateful if you would make it
available to us. If you would like to ``encode'' the grammar first such
that it can no longer be used for language recognition (e.g. by deleting
the parameters/features or giving the nonterminals dummy names) then it
would be fine with us.

Please mail to markjan@cs.kun.nl.

Mark-Jan Nederhof and Janos J. Sarbo
University of Nijmegen
The Netherlands
[I thought people gave up trying to parse natural language using LR methods
in about 1966. Does anyone still use transformational grammars, a la COMIT?

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.