Development of a rule-based Arabic numerals to Yorùbá translator

Number conversion and normalisation are among major pre-processing steps required by a text-to-speech system to enhance itsaccuracy and increase its application domain. This study presents a consistent rule-based system for translating cardinal numeralsinto their correct Yorùbá lexical forms, with the explicit goal of ensuring unambiguous reverse translation from Yorùbá text back todigits. Inability to handle large numbers, produce consistent and precise translation required for reliable reverse translation withoutcomplicating the model are the limitations of existing systems. The paper proposed a hierarchical rule-based method thatdecomposes input numerals (English text or digits) into their constituent vigesimal place values (20, 200, 2000, and 20,000). Variousmorphological operations such as the multiplicative, midpoint, additive and subtractive rules were applied to generate the Yorùbáequivalence. The model was tested on a randomised set of numerals ranging from zero to trillions. Sample Yorùbá outputs werepresented to human linguistic experts and local speakers, who were tasked with providing the corresponding digit forms. The systemachieved 100% accuracy in reverse translation on an input size of 5 million randomly generated numerical values within the rangeof 1 and 400 million, with an average running time of 70 seconds per 1 million numbers. The study presents a Yorùbá numeraltranslation tool optimised to handle large numbers and maintain reverse translation fidelity. The resulting model serves as animportant component for text-to-speech normalisation, and a tool for machine translation pipelines for Yorùbá language, therebycontributing to both technological advancement and linguistic preservation.