LINGUIST List 8.503

Sat Apr 12 1997

Disc: Parsers

Editor for this issue: T. Daniel Seely <seelylinguistlist.org>

Directory

1. Stefan Mueller (I), Parsers, Re: 8.65

Message 1: Parsers, Re: 8.65

MODERATOR'S NOTE: We would like to apologize for the long
delay in posting this message; it is now somewhat outdated,
but in fairness to all, we wanted to be sure it was distributed. 


- - 

Philip A. Bralich, Ph.D. and Derek Bickerton (8.65) claim they
have the best parser in the world because they didn't find any other
comparable parser in the web. If they would have done some web-search,
they would have found several systems which are online more than two
years by now.

The authors cite the NLP Software Registary maintained by the DFKI
(They gave the wrong address by the way. The correct one is:
http://cl-www.dfki.uni-sb.de/cl/registry/
)
If they would have looked through it more carefully they would have
found links to the Babel-System. The system is online since December
1994!
It was developed by me at the Chair for Computational Linguistics at the
Humboldt-University Berlin and is now to be found at:

http://cl-www.dfki.uni-sb.de/~stefan/Babel/e_babel.html

Part of the system is a fully documented (370+ pages) HPSG fragment of
German (not some mysterious top secret syntactic theory).
There is a Java interface to the web. So you get syntax trees as a
result for an analysis, the nodes of which you can expand to feature
structures that contain all available information about the input
phrase.

As far as the coverage of the grammar is concerned you may look
at http://cl-www.dfki.uni-sb.de/~stefan/Babel/java_phaenomene.html

Some real world examples taken from the Esprit corpus you can find under
http://cl-www.dfki.uni-sb.de/~stefan/Babel/Fast/esprit-laeuft-result.html


Apart from that there are many other parsers and services out in the
web. For instance there are some MT systems out there.
I remember one that translates single sentences and another one that
even translates web pages. Both systems are large scale (as far as I
could tell) and efficient. I don't know the exact URLs anymore and I
don't want to bother searching them. This would have been the job of
Philip A. Bralich, Ph.D. and Derek Bickerton.

Apart from this I found it a little bit strange how Philip A. Bralich,
Ph.D. and Derek Bickerton defined the criteria for a good parser.
Those criteria you mentioned seem to be just the features of your
parser. I could imagine a lot of other criteria but in the end it all
depends on the application you have to write the parser for.

Stefan Mueller
- 


- 
DFKI GmbH Tel.: (+49 - 681) 302 - 5295
Stuhlsatzenhausweg 3 Fax: (+49 - 681) 302 - 5338
D-66123 Saarbruecken http://cl-www.dfki.uni-sb.de/~stefan/

http://cl-www.dfki.uni-sb.de/~stefan/Babel/Interaktiv/Babajava/