Gitari T. M., Kinyori S., Mwathi G., Too E. C. and Ireri H. K.2025-04-102025-04-102023Gitari T. M., Kinyori S., Mwathi G., Too E. C. And Ireri H. K. (2023). Deep Learning in African Languages Translation-A Review. In: Isutsa, D. K. (Ed.). Proceedings Of the Chuka University 9th Annual International Research Conference Held in Chuka University, Chuka, Kenya From 24th To 25th November, 2022. 396-408 Pp.https://repository.chuka.ac.ke/handle/123456789/17670tmunene@chuka.ac.ke; dgmwathi@chuka.ac.ke; hkirimi@chuka.ac.keMachine translation (MT) is the use of computers to automatically translate one language to another. Africa has approximately 2,000 spoken languages, however, only 30 African languages have been machine translated. The main technical factor for the low-rate adoption of MT in Africa is the poor translation accuracy of existing machine translators from one African language into English. Currently, there are two approaches to MT in Africa. The first approach is the classical approach; this approach utilizes the direct mapping of input texts to produce a translated output. Examples of classical MT approaches include: statistical-based machine translators (SBMT), rule-based machine translators (RBMT) and hybridized machine translators (HMT). Classical approaches are the most widely adopted MT approach for African languages, the main reason for the wide adoption is the low cost of computing power in utilizing classical approach. However, classical approach has high-levels of inaccuracy due to language structures differentiation. The second approach is the use of Deep learning (DL) MT. Deep learning MT is a field in artificial intelligence concerned with the application of artificial neural networks to mimic the human brain learning process in language translation. Deep learning MT has the advantage of understanding phrases, complex sentence structures, and even slang when compared to classical MT approach. Deep learning has produced results 60-90% more accurate than the classical approach in translating structured languages such as French into English. However, DL has shortcomings in MT, including, high-costs of training and evaluating models, and, DL is data intensive. These review aims to analyze the current status of machine translation approaches in Africa and provide an output recommendation for universalizing applicable MT in African languages translation. The results of these review will be in both graphical and tabular format.enMachine TranslationClassical ApproachDeep LearningArtificial Neural NetworksArticle