Universal Dependencies for Persian

Universal Dependencies for Persian *Mojgan Seraji, **Filip Ginter, *Joakim Nivre *Uppsala University, Department of Linguistics and Philology, Sweden **University of Turku, Department of Information Technology, Finland *firstname.lastname@lingfil.uu.se **figint@utu.fi Abstract The Persian Universal Dependency Treebank (Persian UD) is a recent effort of treebanking Persian with Universal Dependencies (UD), an ongoing project that designs unified and cross-linguistically valid grammatical representations including part-of-speech tags, morphological features, and dependency relations. The Persian UD is the converted version of the Uppsala Persian Dependency Treebank (UPDT) to the universal dependencies framework and consists of nearly 6,000 sentences and 152,871 word tokens with an average sentence length of 25 words. In addition to the universal dependencies syntactic annotation guidelines, the two treebanks differ in tokenization. All words containing unsegmented clitics (pronominal and copula clitics) annotated with complex labels in the UPDT have been separated from the clitics and appear with distinct labels in the Persian UD. The treebank has its original syntactic annota- tion scheme based on Stanford Typed Dependencies. In this paper, we present the approaches taken in the development of the Persian UD. Keywords: Universal Dependencies, Persian, Treebank 1. Introduction In the past decade, the development of numerous depen- dency parsers for different languages has frequently been benefited by the use of syntactically annotated resources, or treebanks (B¨ohmov´a et al., 2003; Haverinen et al., 2010; Kromann, 2003; Foth et al., 2014; Seraji et al., 2015; Vincze et al., 2010). However, treebanks only exist for a small number of languages, and considering the number of 7,000+ languages in the world,1 a large number of lan- guages still lack treebanks. Due to the diverse typologies and grammatical structures that exist across languages, treebanks are created with dif- ferent annotation schemes. These annotation variations can further be explained by different linguistic theories and the syntactic annotations that treebank developers select based on their own preferences (Nivre, 2015). These dissimilari- ties in annotation schemes often have an impact on syntac- tic parsers, which means that the results are not comparable across languages. McDonald et al. (2013) enumerate sev- eral issues for natural language parsing when treebanks are labeled with different annotation schemes. This has brought many researchers and developers in nat- ural language processing to the conclusion that having a common standard and cross-linguistically valid annotation scheme would favor parsing research. Recently, there have been a number of initiatives for developing data sets with cross-linguistically consistent annotation scheme for mor- phological and syntactic structures. These efforts have re- sulted in the emergence of the Stanford Typed Dependen- cies Representation (de Marneffe et al., 2006; de Marneffe and Manning, 2008), the Google Universal Part-of-Speech Tagset (Petrov et al., 2012), and Interset interlingua for morphosyntactic features used in the HamleDT treebank collection (Zeman, 2008; Zeman et al., 2012). The most recent effort is the Universal Dependencies (UD), which more or less combine all the earlier efforts in this regard. In 1http://www.bbc.co.uk/languages/guide/languages.shtml this paper, we present how we adapt the Universal Depen- dencies to Persian by converting the Uppsala Persian De- pendency …

Content truncated to 500 words