Breaking Language Barriers: India's AI Push for Multilingual Accessibility

In the diverse linguistic landscape of India, new technologies like AI-driven language translation are revolutionizing communication for millions, making vital services more accessible.

Vineet Sawant, a Mumbai delivery driver, has experienced first-hand how AI translation tech can transform work dynamics. For someone whose command of English reads limited, the solution lay in a collaboration between his employer, Zepto, and Reverie Language Technologies. “When I first started, it was tough to decipher delivery instructions,” he shared while navigating the bustling streets of Mumbai. “Everything was in English. Now, I receive instructions in Marathi, my native language.”

In a nation known for its 22 official languages and a plethora of dialects, communication barriers have often hindered progress, particularly for those in employment sectors reliant on clear instructions. As more organizations adopt AI-driven translation services, millions are finding their roles and responsibilities made simpler, as Mr. Sawant's experience demonstrates. This shift is timely, as relevant AI systems must evolve to keep pace with a surging demand for multilingual capabilities, particularly given the rise of platforms like ChatGPT.

Professor Pushpak Bhattacharyya from IIT Mumbai elucidates the importance of creating quality datasets that allows AI to understand and speak Indian languages. “The data available tends to be coarse and often lacks refinement,” he noted, pointing out that many linguistic minority groups remain underrepresented in digital environments. The limited digitization of tribal and regional dialects exacerbates this issue, making AI training a significant challenge across the board.

Reverie Language Technologies aims to bridge this gap, focusing its AI translation technology on various Indian companies. Co-founder Vivekananda Pani cautioned that while these advancements promise improved communication, there's a risk that less common dialects may be left behind in favor of more dominant languages.

Responding to this, Professor Bhattacharyya has become involved with Bhashini, an expansive government project initiated in 2022 designed to build high-quality datasets for training AI. With over 350 AI-driven language models currently serving more than a billion tasks, Bhashini collaborates with various government departments to enhance public service accessibility. As Amitabh Nag, CEO of Digital India for Bhashini Division, stated, “We are creating India-specific AI models to safeguard linguistic and cultural representation.”

The vision for Bhashini is ambitious; it targets making voice-enabled government services and information systems accessible to rural communities in their indigenous languages within a few years.

Amidst these technological advancements, researchers like Kshitij Jadhav from the Koita Centre for Digital Health are also focusing their efforts on developing AI systems in healthcare. Jadhav is particularly interested in designing AI solutions that help individuals quit smoking, ensuring that the guidance provided is tailored to the linguistic and emotional needs of users.

As Vineet Sawant continues to thrive on the streets of Mumbai, delivering up to 30 parcels daily, he expresses optimism about the future. "It makes us feel like we belong. Not everyone knows English. When the app works in our language, it builds our confidence, allowing us to deliver more effectively." With the momentum of AI-driven language translation, more lives like Sawant's can be positively impacted, shaping a more inclusive future for India's diverse workforce.