The FraCaS GF Treebank

- --- - - - - - - - - - - - -

Authors:	Peter Ljunglöf - Magdalena Siverbo
Version:	0.2
Date:	2012-01-27
Organization:	Centre for Language Technology, University of Gothenburg
Copyright:	Distributed under GNU GPL v3, see COPYING.txt for details

1. Introduction

This is the FraCaS Treebank, developed and maintained by -the Centre for Language Technolgy at University of Gothenburg:

-http://www.clt.gu.se/

The treebank is part of the CLT Toolkit, a set of state-of-the-art -open source Language Technology tools and accompanying linguistic -resources. The different parts of the toolkit, including the -FraCaS Treebank, can be downloaded from:

-http://www.clt.gu.se/clt-toolkit

The treebank is built upon the FraCaS textual inference problem set, -which was built in the mid 1990’s by the FraCaS project, a large -collaboration aimed at developing resources and theories for -computational semantics. This test set was later modified and -converted to XML by Bill MacCartney:

-http://www-nlp.stanford.edu/~wcmac/downloads/fracas.xml

It is this modified version that has been used in this treebank. -The corpus consists of 346 problems each containing one or more -statements and one yes/no-question (except for four problems, where -there is no question). The total number of sentences in the corpus is -1220, but since some of them are repeated in several problems, there -are in total 874 unique sentences.

2. Description

The treebank is created in Grammatical Framework (GF), using its -multilingual Resource Grammar as backend grammar. Currently the -treebank is bilingual, with an English and a Swedish lexicon.

More information about GF, including installation instructions, -can be found at:

-http://www.grammaticalframework.org/

The treebank is also distributed in XML and Prolog formats, -for people that have no interest in learning GF. Note however -that the syntactical constructions come from the GF resource grammar.

3. Download and installation

The full distribution can be downloaded from -dist/FraCaSBank-0.2.zip.

The Prolog and XML treebanks are already generated, so to use these -you don't need anything else. But if you want to work with the GF -source files, you need a GF installation including the Resource Grammar.

4. Contents

a) Documentation

The documentation is located in the doc directory:

FraCaSBank-report.{pdf,lyx,bib}:: A technical report describing the treebank, together with -the LyX and -BibTeX source files. -The PDF version can be read here.

b) GF source files

The grammar sources are located in the src directory:

Additions*.gf: Generic additions to the GF Resource Grammar.
FraCaS*.gf: Grammatical constructions specific to the FraCaS domain.
FraCaSLex*.gf: The lexical items in the FraCaS treebank.
FraCaSBank*.gf: The actual treebank. -The file FraCaSBankOriginal.gf contains the original treebank sentences. -The file FraCaSBankI.gf contains the language-independent abstract syntax trees.

c) Other files

Makefile, build_fracasbank.py: Files for automatically generating the XML and Prolog treebank.
build/FraCaSBank.{xml,pl}: The automatically generated -XML treebank and -Prolog treebank.
dist/FraCaSBank*.zip: All files collected in a zip file.
README.txt, COPYING.txt:: The source of this README file, and GNU GPL 3 licensing information.