Estonian Dependency Treebank: from Constraint Grammar tagset to Universal Dependencies

Estonian Dependency Treebank: from Constraint Grammar Tagset to Universal Dependencies Kadri Muischnek, Kaili Müürisep, Tiina Puolakainen University of Tartu Estonia Email: kadri.muischnek@ut.ee, kaili.muurisep@ut.ee, tiina.puolakainen@ut.ee Abstract This paper presents the first version of Estonian Universal Dependencies Treebank which has been semi-automatically acquired from Estonian Dependency Treebank and comprises ca 400,000 words (ca 30,000 sentences) representing the genres of fiction, newspapers and scientific writing. Article analyses the differences between two annotation schemes and the conversion procedure to Universal Dependencies format. The conversion has been conducted by manually created Constraint Grammar transfer rules. As the rules enable to consider unbounded context, include lexical information and both flat and tree structure features at the same time, the method has proved to be reliable and flexible enough to handle most of transformations. The automatic conversion procedure achieved LAS 95.2%, UAS 96.3% and LA 98.4%. If punctuation marks were excluded from the calculations, we observed LAS 96.4%, UAS 97.7% and LA 98.2%. Still the refinement of the guidelines and methodology is needed in order to re-annotate some syntactic phenomena, e.g. inter-clausal relations. Although automatic rules usually make quite a good guess even in obscure conditions, some relations should be checked and annotated manually after the main conversion. Keywords: universal dependencies, dependency grammar, Estonian language 1. Introduction Estonian Dependency Treebank (EDT) is a dependency- annotated treebank of written Estonian1. It comprises ca 400,000 words (ca 30,000 sentences); the texts represent the genres of fiction, newspapers and scientific writing (Muischnek et al., 2014b). The treebank is annotated semi-manually for lemma, part of speech, morphological categories, syntactic functions and dependency relations. The original morphological tagset2 is a language-specific local standard, whereas the set of syntactic relations is based on Constraint Grammar (Karlsson, 1990; Karlsson et al., 1995) and coding of dependency relations is based on an expansion of Constraint Grammar (Bick & Didriksen, 2015). Several large projects with the aim of creating treebanks in multiple languages have emerged during the the past few years. These projects try to introduce homogeneous or universal dependency annotations (McDonald et al., 2013) as such annotation enables building better syntax-based machine translation systems and other language technology applications. The general aim of Universal Dependencies3 (UD) project is to provide a cross-linguistically and typologically 1 The EDT is freely available at https://github.com/EstSyntax/EDT 2 The documentation is available at http://www.cl.ut.ee/korpused/morfliides/seletus.php? lang=en consistent inventory of categories and guidelines in order to facilitate consistent annotation of similar constructions across languages and thus promote cross-linguistically relevant methods and tools in natural language processing (e.g. Nivre, 2015). There are currently available Universal Dependencies for more than 40 languages. The list of UD-related publications4 contains articles about Finnish (Pyysalo et al., 2015) and Swedish (Nivre, 2014) Universal Dependencies. It seems that the conversion of Finnish (Turku Dependency) Treebank and Swedish Treebank to Universal Dependencies was an easier or at least less fuzzy task than converting the Constraint Grammar based EDT annotation. The original Turku Dependency Treebank annotation was derived from Stanford Dependencies, which is also the predecessor of UD. Although the …

Content truncated to 500 words