Digital Corpora for Igbo and Edo: Imperatives for NLP and ML Development

Abstract

The world is fast leveraging on Natural Language Processing (NLP) and Machine Learning (ML) to achieve enormous breakthroughs in language matters, evidenced in what are now possible in the use of different languages by digital appliances and robots, and translation abilities of AI tools. The advancement of NLP and ML that has brought these achievements is possible through the availability of elaborate, organized and easily accessible corpora. Such breakthroughs are still restricted to few languages like English, German, Spanish and Chinese due to lack of digital corpora which has specifically denied many African languages benefit in terms of NLP and ML. Languages with enormous and retrievable digital corpora such as websites, textbooks, literary materials and metadata, can easily participate in NLP and ML activities, unlike languages without such corpora. There raises the need for speakers and linguists interested especially Nigerian languages, to make conscious efforts at developing the digital corpora of their object languages. This paper advocates for elaborate, organized, well archived and easily accessible corpora for the Igbo and Edo languages to augment the few limited ones available in the languages such as Igbo Text Analyzer, IgbTC (Igbo Tagged Corpus), and AfriDataHub's Igbo Repository; Edo Language Dictionary, OTranslate, and Speak and Write Edo Language. The availability of robust digital corpora will ensure more elaborate language research, detailed linguistic analysis, extensive language preservation efforts, and language modelling for machine translation, text-to-speech and speech-to-text creation. It will further enhance the more active participation of these languages for developments in the digital space.