International Research Journal of Education and Technology

Recent advances in natural language processing (NLP), especially in retrieval augmented generation (RAG), have greatly increased response accuracy through integrated information retrieval and information processing but so this technology mainly focuses on widely spoken languages ​​like English, leaving languages ​​like Tamil less available. This study aims to improve language technology for Tamil speakers by developing a specialized RAG model for Tamil. This model was developed to address the unique grammatical and cultural aspects of Tamil. By training the Tamil text, the model can give more accurate and relevant answers. It performed well in simple queries but had trouble with words with multiple meanings, regional accents and old Tamil. Even if possible, more changes are needed to address complex questions. Finally, this study shows that the Tamil-specific RAG model can make language technology more favorable for Tamil speakers, so that the responses are authentic and culturally appropriate and further developments enable the model to work well.

Key Words:  Tamil Language, Retrieval Augmented Generation (RAG), Natural Language Processing (NLP), Language Model, Information Retrieval, Text Generation, Machine Learning.