AI-Based Multilingual Lecture Translation For Higher Education: A Solution For First Year Students

This project addresses a major challenge faced by first-year Computer Science students in South Africa, particularly at the University of the Western Cape (UWC), where English is the primary language of instruction, despite most students speaking other languages as their first language. The technical and abstract nature of Computer Science further intensifies this language barrier, making lectures difficult to understand. To address this issue, the project develops an AI-based multilingual lecture translation system that provides real-time translation of lectures into isiZulu, isiXhosa and Afrikaans. The system is delivered through a web-based interface and integrates three key technologies: Automatic Speech Recognition (ASR), Neural Machine Translation (NMT) and Text-to-Speech (TTS), enabling accurate and natural sounding translations. The results indicate that the system significantly improves translation accuracy, enhances student understanding as well as academic performance. While there are various limitations such as the support for only 3 languages or one directional translation only, the solution offers an inclusive, scalable and cost-effective approach to reducing language barriers and promoting accessibility in higher education for first year students.