ParGramBank: The ParGram Parallel Treebank

Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 550–560, Sofia, Bulgaria, August 4-9 2013. c⃝2013 Association for Computational Linguistics ParGramBank: The ParGram Parallel Treebank Sebastian Sulger and Miriam Butt University of Konstanz, Germany {sebastian.sulger|miriam.butt}@uni-konstanz.de Tracy Holloway King eBay Inc., USA tracyking@ebay.com Paul Meurer Uni Research AS, Norway paul.meurer@uni.no Tibor Laczk´o and Gy¨orgy R´akosi University of Debrecen, Hungary {laczko.tibor|rakosi.gyorgy}@arts.unideb.hu Cheikh Bamba Dione and Helge Dyvik and Victoria Ros´en and Koenraad De Smedt University of Bergen, Norway dione.bamba@lle.uib.no, {dyvik|victoria|desmedt}@uib.no Agnieszka Patejuk Polish Academy of Sciences aep@ipipan.waw.pl ¨Ozlem C¸ etino˘glu University of Stuttgart, Germany ozlem@ims.uni-stuttgart.de I Wayan Arka* and Meladel Mistica+ *Australian National University and Udayana University, Indonesia +Australian National University wayan.arka@anu.edu.au, meladel.mistica@gmail.com Abstract This paper discusses the construction of a parallel treebank currently involving ten languages from six language families. The treebank is based on deep LFG (Lexical- Functional Grammar) grammars that were developed within the framework of the ParGram (Parallel Grammar) effort. The grammars produce output that is maxi- mally parallelized across languages and language families. This output forms the basis of a parallel treebank covering a diverse set of phenomena. The treebank is publicly available via the INESS tree- banking environment, which also allows for the alignment of language pairs. We thus present a unique, multilayered paral- lel treebank that represents more and dif- ferent types of languages than are avail- able in other treebanks, that represents deep linguistic knowledge and that allows for the alignment of sentences at sev- eral levels: dependency structures, con- stituency structures and POS information. 1 Introduction This paper discusses the construction of a parallel treebank currently involving ten languages that represent several different language families, in- cluding non-Indo-European. The treebank is based on the output of individual deep LFG (Lexical- Functional Grammar) grammars that were deve- loped independently at different sites but within the overall framework of ParGram (the Parallel Grammar project) (Butt et al., 1999a; Butt et al., 2002). The aim of ParGram is to produce deep, wide coverage grammars for a variety of lan- guages. Deep grammars provide detailed syntactic analysis, encode grammatical functions as well as 550 other grammatical features such as tense or aspect, and are linguistically well-motivated. The Par- Gram grammars are couched within the linguis- tic framework of LFG (Bresnan, 2001; Dalrymple, 2001) and are constructed with a set of grammati- cal features that have been commonly agreed upon within the ParGram group. ParGram grammars are implemented using XLE, an efficient, industrial- strength grammar development platform that in- cludes a parser, a generator and a transfer sys- tem (Crouch et al., 2012). XLE has been devel- oped in close collaboration with the ParGram project. Over the years, ParGram has continu- ously grown and includes grammars for Ara- bic, Chinese, English, French, German, Georgian, Hungarian, Indonesian, Irish, Japanese, Mala- gasy, Murrinh-Patha, Norwegian, Polish, Spanish, Tigrinya, Turkish, Urdu, Welsh and Wolof. ParGram grammars produce output that has been parallelized maximally across languages ac- cording to a set of commonly agreed upon uni- versal proto-type analyses …

Content truncated to 500 words