|prefix.html||23-Oct-2012 14:59||1.1K||FraCaS treebank visual>|
|Organization:||Centre for Language Technology, University of Gothenburg|
|Copyright:||Distributed under GNU GPL v3, see COPYING.txt for details|
This is the FraCaS Treebank, developed and maintained by the Centre for Language Technolgy at University of Gothenburg:
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:
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:
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.
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:
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.
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.
The documentation is located in the doc directory:
The grammar sources are located in the src directory: