The UD Scheme and the Test Treebank of Georgian

• In Event: Current Issues in Eurasian Scholarship

Abstract

One of the most crucial Natural Language Processing (NLP) task is associated with the universality-driven development of language resources for different languages (e.g. Universal Dependencies (UD), UniMorph, PARSEME etc.). The Universal Dependencies (UD) community has released a huge number of Treebanks with consistent cross-linguistic grammatical annotation, but this resource lacks an appropriately annotated Georgian Treebank. This is because Georgian suffers from data scarceness, i.e. the amount of data to train NLP tools is not sufficient, but also because the tools developed for other languages cannot be easily adopted in case of Georgian, due to the differences between morphosyntactic annotation schemes. Thus, the aim of the paper is to describe the universal dependencies scheme for Georgian, which ensures the compatibility of the annotation schemes cross-linguistically and enriches the universal dependencies with Georgian data and compiles the Test Treebank for Georgian

Author

• Irina Lobzhanidze, Ilia State U (Georgia)