Tree
Fragments

XMG-based XTAG

Core Tree-Adjoining MetaGrammar for English

Introduction

This project focused on developing a computational grammar for English within the Tree-Adjoining Grammar (TAG) formalism. The implementation of the grammar was based on the linguistic decisions made in the large-coverage English TAG (see XTAG). The current project was carried out within the eXtensible MetaGrammar (XMG) framework (developed at LORIA, Nancy) which is specially designed for factorising strongly lexicalised tree-based grammars and grammars equipped with a syntax/semantics interface. The XMG description language was built upon the observation that a fine-grained grammar can be modelled on a higher level of abstraction by defining as tree descriptions a set of reusable tree fragments and declaring rules for their composition. A good factorisation was reached by organising the tree fragments in a hierarchy and by combining them using two logical operations: conjunction for specifying shared structures and disjunction for representing possible alternatives.


Organisation of the Grammar

The grammar was implemented at four consecutive levels of abstraction so that each level builds on the grammatical information defined at the previous level. This modular architecture was inspired by [Candito][1] and it also served as a main inspiration for the development of the core TAG for French [Crabbé] [2]. Starting from the highest level, the grammar was organised as follows:

  • Verb Families: set of trees associated with verbs of a certain syntactic class. For instance, n0Vn1 family defines the set of trees associated with transitive verbs. The code of the implemented families along with documentation can be found here.
  • Diathesis: set of trees associated with a given (verb family, diathesis) pair. Here we capture the active/passive alternations within a verb family. The code of the implemented diathesis alternations for each tree family along with documentation can be found here.
  • Syntactic Functions: set of tree fragments associated with each possible syntactic function. The code of the implemented syntactic functions along with documentation can be found here.
  • Tree Fragments: the basic building blocks of the grammar defined as tree descriptions and organised in an inheritance hierarchy. The code of the implemented tree fragments along with documentation can be found here.

Grammar

The grammar is organised in six .mg files as follows:


Lexicon & Test Suite

The lexicon and the test suite can be found here:


Selected Bibliography

[1]
Marie-Hélène Candito.Représentation modulaire et paramétrable de grammaires électroniques lexicalisées : application au français et à l'italien. Université Paris 7, 1999. PhD thesis. [ bib ]
[2]
B. Crabbé and D. Duchier. Metagrammar Redux. In International Workshop on Constraint Solving and Language Processing - CSLP 2004, Copenhagen, 2004. [ bib ]
[3]
Benoit Crabbé. Représentation informatique de grammaires fortement lexicalisées: Application à la grammaire d'arbres adjoints. PhD thesis, 2005. Ph.D. thesis. [ bib ]
[4]
D. Duchier, J. Le Roux, and Y. Parmentier. The Metagrammar Compiler: An NLP Application with a Multi-paradigm Architecture. In Second International Mozart/Oz Conference (MOZ'2004), Charleroi, 2004. [ bib | http ]
[5]
C. Gardent and L. Kallmeyer. Semantic Construction in FTAG. In Proceedings of the 10th EACL, Budapest, Hungary, 2003. [ bib ]
[6]
C. Gardent and E. Kow. Generating and Selecting Grammatical Paraphrases. Proceedings of the ENLG, Aug 2005. [ bib ]
[7]
C. Gardent and E. Kow. A Symbolic Approach to Near-deterministic Surface Realisation Using Tree-Adjoining Grammar. In ACL'07, Prague, 2007. [ bib ]
[8]
Sylvain Pogodalla. Computing Semantic Representation: Towards ACG Abstract Terms as Derivation Trees. In Proceedings of the Seventh International Workshop on Tree Adjoining Grammar and Related Formalisms (TAG+7), pages 64-71, May 2004. [ bib ]
[9]
XTAG-Research-Group. A lexicalized Tree-Adjoining Grammar for English. Technical Report IRCS-01-03, IRCS, University of Pennsylvania, 2001. [ bib | http ]
[10]
Y. Parmentier and J. Le Roux. XMG: an Extensible Metagrammatical Framework. In Student session of the 17th European Summer School in Logic, Language and Information,, Aug 2005. [ bib ]

This file has been generated by bibtex2html 1.88.