Published Paper Details:

India's linguistic diversity, featuring 22 recognized languages and numerous local dialects, poses a major obstacle to digital inclusion and intercultural communication. Although global translation tools excel in standardized languages, they often face challenges with the subtleties of regional expressions, local sayings, and culturally unique terms. This document suggests creating a localized translation app aimed at promoting genuine interaction among speakers of various Indian languages. Employing advanced Large Language Models (LLMs) and Natural Language Processing (NLP) methods, the study concentrates on recognizing and translating "non-literal" regional phrases that conventional systems frequently misread. The approach includes refining multilingual models using selected datasets of informal language and utilizing Retrieval-Augmented Generation (RAG) to deliver context-sensitive translations. Initial findings indicate that emphasizing phrase-level cultural context greatly enhances user understanding and learning results in contrast to literal translation. This study aids in the democratization of digital resources in India, providing a scalable approach to close the communication divide in education, commerce, and social engagement.