Transformer-Based Sinhala Java Programming Learning Tool
D. I. De Silva, K. S. N. Athukorala · 2024
Programming is a fundamental aspect of computer science, yet it presents significant language challenges for non-native English-speaking students. This study focuses on developing a Java programming assistance tool tailored for Sri Lankan novice programmers who are native Sinhala speakers. The primary challenge addressed is the difficulty these students face when using English-based programming resources. To overcome this, the study introduces an innovative assistance tool that integrates a custom transformer-based Sinhala- English translation model with advanced AI technology, specifically ChatGPT, for real-time code generation and explanations in Sinhala. The tool is designed to accept queries in Sinhala and provide relevant Java snippets or explanations of Java code in Sinhala. The system architecture includes a user-friendly front-end built on Flask, a robust back-end API, and an accurate translation model with an 83 % accuracy rate. Usability tests with ten beginner Sinhala-speaking programmers demonstrated significant improvements in their understanding and retention of programming concepts, with an average increase in test scores by 25%. This tool enhances programming education accessibility for Sinhala speakers, promoting a more inclusive global coding community.