The Treebank of Vedic Sanskrit

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 5137–5146 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 5137 The Treebank of Vedic Sanskrit Oliver Hellwig, Salvatore Scarlata, Elia Ackermann, Paul Widmer Department of Comparative Language Science, University of Zurich Abstract This paper introduces the first treebank of Vedic Sanskrit, a morphologically rich ancient Indian language that is of central importance for linguistic and historical research. The selection of the 4,000 sentences contained in this treebank reflects the development of metrical and prose texts over a period of 600 years. We discuss how these sentences are annotated in the Universal Dependencies scheme and which syntactic constructions required special attention. In addition, we describe a syntactic labeler based on neural networks that supports the initial annotation of the treebank, and whose evaluation can be helpful for setting up a full syntactic parser of Vedic Sanskrit. Keywords: Treebank, Vedic Sanskrit, Universal dependencies 1. Introduction Vedic Sanskrit (VS) is an ancient Indo-Aryan language, one of the oldest transmitted Indo-European languages and the precursor of Classical Sanskrit.1 The large corpus of Vedic poetry and prose is important for reconstructing the early linguistic history of Indo-European and as a source for socio-cultural developments in South Asia during the sec- ond and first millenia BCE. The transmission of VS starts with its most famous text, the R. gveda (R. V), composed pre- sumably in the 2nd millenium BCE and comprising reli- gious hymns. While the second oldest Vedic text, the Athar- vaveda, focusses on royal and medicinal rites, the bulk of the following Vedic literature discusses the Vedic ritual and ends, at around 500-300 BCE, with texts that demarcate the transition to the early Buddhist culture (Witzel, 1997; Witzel, 2009). Inspired in part by the famous Vedic grammarian P¯an.ini, Indo-European and Vedistic research have studied the con- tent and the linguistic structure of VS for over 150 years. Most of these studies are, however, based on small sub- corpora of VS, often only on parts of the R. V, and there- fore do not cover but a small part of the sociolinguistic and spatiotemporal variation actually encountered in VS. Moreover, the results of previous studies are often difficult to reproduce, when larger, more diverse text samples are considered. The composition of the Vedic treebank (VTB) introduced in this paper is motivated by the need for a re- source that can be used for data-driven, quantitatively ro- bust diachronic and synchronic investigations of linguistic phenomena in, and starting with, the oldest layers of VS. One topic we are particularly interested in is the question whether the early metrical texts show a higher degree of non-configurationality than the later prose texts (Gillon, 1996; Kulkarni et al., 2015; Rein¨ohl, 2016). Consider the passage R. gveda 1.51.5c, where the discontinuous elements of the NP piproh. purah. ‘the strongholds of Pipru (a man)’ are printed in bold: 1Abbreviations used in this paper: AB: Aitareya Br¯ahman.a, MS: Maitr¯ayan.¯ı Sam. hit¯a, R. V: R. …

Content truncated to 500 words