Multi-layer Annotation of the Rigveda

Multi-layer Annotation of the R. gveda Oliver Hellwig1 , Heinrich Hettrich2 , Ashutosh Modi3 , Manfred Pinkal3 1SFB 991, Düsseldorf University, ohellwig@phil-fak.uni-duesseldorf.de 2Würzburg University, heinrich.hettrich@uni-wuerzburg.de 3Department of Language Science and Technology, Saarland University, {ashutosh,pinkal}@coli.uni-saarland.de Abstract The paper introduces a multi-level annotation of the R. GVEDA, a fundamental Sanskrit text composed in the 2. millenium BCE that is important for South-Asian and Indo-European linguistics, as well as Cultural Studies. We describe the individual annotation levels, including phonetics, morphology, lexicon, and syntax, and show how these different levels of annotation are merged to create a novel annotated corpus of Vedic Sanskrit. Vedic Sanskrit is a complex, but computationally under-resourced language. Therefore, creating this resource required considerable domain adaptation of existing computational tools, which is discussed in this paper. Because parts of the annotations are selective, we propose a bi-directional LSTM based sequential model to supplement missing verb-argument links. Keywords: Sanskrit, multi-layer corpus, verb-argument structures, R. igveda 1. Introduction In this paper, we introduce a multi-layer annotation of the complete R. GVEDA (R. V). The R. V is central for research in Indo-European linguistics, because it is the oldest sam- ple of this language family for which a sizeable text corpus has been transmitted (Witzel, 1995). The Vedic core corpus consists of four collections of hymns called Vedas (“know- ledge”), which deal mainly with the worship of the Vedic pantheon and details of the ritual. The R. V is the oldest among these four collections. It may have been composed around 1,500 BCE, has been transmitted orally for at least two millenia, and has, in spite of its age, remained a founda- tional text for the religious, cultural, and linguistic history of South Asia. Ideas and actors mentioned in the R. V are constantly referred to in later texts produced on the Indian subcontinent (Gonda, 1975). The corpus presented in this paper merges the phonetic, morphological, and lexical analyses for each word of the R. V with a verb-argument (VA) annotation that links each verbal form to its main syntactic arguments. The first part of the paper describes how the annotation was performed, and provides a quantitative overview of size and structure of the resulting corpus. The second part of the paper presents a basic argument identification algorithm for Vedic Sanskrit. The VA annotation was created in a linguistic research con- text, but not as a building block of an NLP pipeline. As a consequence, parts of the case semantic information are not encoded explicitely, but need to be supplemented by a human reader. We will use the presented argument identifi- cation algorithm for complementing these missing parts of the VA annotation, and, later on, for annotating other cen- tral texts of the Vedic corpus. This paper makes two important contributions. First, it in- troduces a novel resource for Sanskrit with deep linguistic annotations, ranging from the phonetic up to the syntactic level. The full annotated corpus is available via https:// git.adwmainz.net/open/rigveda under the Cre- ative Commons Attribution 4.0 International Public Li- cense. …

Content truncated to 500 words