Digitizing Old Nubian Dictionary: Optical Character Recognition for Multi-Lingual and Multi-Script Text from Medieval Africa

This paper explored the application of the Transkribus software, a Handwritten Text Recognition (HTR) and Optical Character Recognition (OCR) tool, to multi-lingual and multi-script text from Medieval Nubia in Africa. Transkribus leverages machine learning algorithms using Recurrent Neural Networks (RNNs) to enhance accuracy, specifically targeting mixed publications in Old Nubian, Coptic, Greek, Arabic, and English. The output of the Transkribus processing facilitates the analysis, search, and translation of historical documents, ensuring their preservation and accessibility for future researchers and scholars worldwide. We aim to underscore the growing importance of accurate OCR technology in safeguarding and disseminating invaluable historical resources while highlighting Transkribus’s role in this process.