📄 Universal Dependencies: Principles and Tools (tutorial)

Summary

This work is not an experimental paper but a short abstract of a tutorial for the CLiC-it 2021 conference. It introduces the audience to Universal Dependencies (UD). UD is both an international community of researchers and an open collection of treebanks. In these treebanks, sentences are annotated morphologically and syntactically according to a unified scheme. At the time, the collection covered more than a hundred languages. The tutorial consists of two parts. First, the author briefly explains the basic principles of UD annotation. He then moves on to practice: he presents the data themselves and the software for working with them. These include parsers, batch processing tools, search engines and syntactic tree viewers. The text contains no new models, evaluations or metrics such as LAS. The value of the work lies in providing an introductory overview of a resource that has become a standard in multilingual syntactic parsing. The author emphasizes that UD is suitable for a wide range of linguistic research: from individual constructions in a single language to comparative typology, the description of endangered languages and the study of historical language change.

Key points

• Type of work: tutorial abstract at CLiC-it 2021. It contains no experimental results or quality figures.
• UD combines an international community project and a set of treebanks with morphosyntactic annotation. As of 2021, the set covered more than 100 languages.
• Applications: the grammar of individual languages, linguistic typology, documentation of endangered languages, diachronic studies.
• Tools covered in the tutorial: parsers, batch processors, search engines, viewers.
• The author is Daniel Zeman of ÚFAL (Charles University, Prague). He is a co-founder and one of the main coordinators of UD.