Expanding the Lexicon of Ge'ez-Based African Languages: Amharic and Tigrinya

arXiv·low signal

Multilingual pretrained language models degrade sharply on low-resource, non-Latin-script languages, and this study tackles that gap for the Ge'ez-script languages Amharic and Tigrinya through a comparative lexicon-expansion approach. It contributes resources and analysis for improving PLM coverage of under-served African languages. Relevant to practitioners working on multilingual NLP and equitable language coverage beyond high-resource scripts.