Yorùbá Dependency Treebank (YTB)

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 5178–5186 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 5178 Yorùbá Dependency Treebank (YTB) Ọlájídé Ishola, Daniel Zeman Charles University, Faculty of Mathematics and Physics, ÚFAL Malostranské náměstí 25, CZ-11800 Praha lajish@tasmag.com, zeman@ufal.mff.cuni.cz Abstract Low-resource languages present enormous NLP opportunities as well as varying degrees of difficulties. The newly released treebank of hand-annotated parts of the Yorùbá Bible provides an avenue for dependency analysis of the Yorùbá language; the application of a new grammar formalism to the language. In this paper, we discuss our choice of Universal Dependencies (UD), important annotation decisions, and results of our parsing experiments. We also highlight future directions for a rapid expansion of the treebank. Keywords: Universal Dependencies, Yorùbá, annotation, treebank, grammar, parsing 1. Introduction Yorùbá is one of the most spoken indigenous African languages with estimated 40 million speakers,1 living mostly in Nigeria, Benin, Togo, and across the At- lantic in Brazil, Cuba, Jamaica and Trinidad. How- ever, there is a distinction between the African and diaspora Yorùbá. Yorùbá is classified as a Niger-Congo language which shares close relations with Itsekiri and Igala (Akinlabí and Adéníyì, 2017). It is one of the three major Nige- rian languages recognised in the National Policy on Education (NPE) document published in 1977 to be used as a medium of instruction for the first 3 years of a child’s primary school education. Yorùbá is also offered as a course of study in some Nigerian tertiary institutions. Yorùbá also has Braille notations; a part of the Nigerian Braille writing system for the visually impaired, coordinated by Braille Advancement Associ- ation of Nigeria (BRAAN) (UNESCO, 2013). The earliest work on Yorùbá appeared in 1819 century when the German linguist Bowdich published a vocab- ulary primer containing the numerals 1-10 (Ogunbiyi, 2003). There were several individual and collabora- tive efforts thereafter to arrive at the contemporary Yorùbá orthography and scholarship. Today, Yorùbá is one of the most documented West African languages (Akinlabí and Adéníyì, 2017). Despite the extensive language resources available in the mass media, film industry, books and rich un- documented oral literature, they remain untapped for open-source annotated data. Hence, Yorùbá is only low-resource from a technological point of view due to lack of readily-available corpora for computational analysis. We describe our work on the creation of the first Yorùbá Treebank (YTB) using the Universal De- pendencies (UD) framework on data sourced from the Yorùbá Bible. 1https://www.ethnologue.com/language/yor 2. Universal Dependencies Despite vast dissimilarities in the languages of the world, there are attempts at formalisms that can be applied to all languages. However, minority languages, especially African languages, do not receive enough at- tention due to lack of annotated data. Beaming search light on “low-resource” languages will help us make better and more universal grammar rules. The Universal Dependencies (UD) project (Nivre et al., 2020) was started to provide a universal inventory of categories/tagsets (allowing language-specific exten- sions where necessary) and guidelines …

Content truncated to 500 words