[Prev][Next][Index]
[xtag-meeting]: This week: Alexandra Kinyon
-
To: xtag-meeting@cis.upenn.edu
-
Subject: [xtag-meeting]: This week: Alexandra Kinyon
-
From: Anoop Sarkar <anoop@linc.cis.upenn.edu>
-
Date: Tue, 11 Jun 2002 13:18:01 EDT
-
Sender: owner-xtag@linc.cis.upenn.edu
XTAG Meeting: 10:30am, 06/13/2002, IRCS Fishbowl.
This week Alexandra Kinyon will talk about XTAG and Metagrammars. Her talk
abstract follows.
Replicating the Xtag grammar for English using a MetaGrammar
------------------------------------------------------------
In this talk, I will describe work in progress which aims at automatically
generating a replication of the Xtag grammar for English, as it is described
in the Xtag technical report. The grammar is currently partially generated
using the MetaGrammar tool described in (Gaiffe and Crabbe, 2002).
I will address the following points, and hope that you can help me find
answers to some of those questions:
- Can a MetaGrammar implementation of Xtag ease grammar maintenance ?
- Can the MetaGrammar be used for Hypertagging in order to obtain high level
syntactic descriptions of lexical items ?
- Can a MetaGrammar approach be coupled with the MetaRule approach presented
in (Prolo 2002) ?
- Can a MetaGrammar representation be helpful for integrating semantics into
Xtag ?
- Can a MetaGrammar representation be of any use for parsing ?
I will also discuss some of the problems encountered while encoding the MG,
such as the problem of exponentiality of the generating algorithm, and the
lack of "standard" format for TAG related tools.
Finally, if there is enough time, I will propose a classification of four
grammar development strategies based on the two distinction:
- "hand-written" versus "automatically acquired" grammars
- "low level of syntactic abstraction" versus "high level of syntactic
abstraction"