In the bustling streets of Mumbai, the growing intersection of technology and language is reshaping the work environment for delivery drivers like Vineet Sawant. As someone who has spent the past two years delivering groceries across the city, Sawant reflects on the challenges he faced due to language barriers. A native Marathi speaker with limited English proficiency, he struggled initially with reading instructions in English. "I would often seek help from others, as I wasn't comfortable deciphering the messages," he explains. Zepto, Sawant's employer, valued efficiency and customer service in their promise of "India's Fastest Online Grocery Delivery," making it imperative to address these language hurdles.
To optimize the delivery process, Zepto collaborated with Reverie Language Technologies, resulting in an AI-driven translation feature on their app that now supports six languages. This technology has empowered drivers like Sawant to receive vital instructions clearly in their preferred language, as he recounts, "It’s all clear now. I don’t have to check again or guess." With 22 official languages and numerous dialects in India, the need for advanced translation technology is crucial for inclusivity in various sectors such as education, governance, and healthcare.
Professor Pushpak Bhattacharyya from IIT Mumbai emphasizes the importance of language technology: "Without tools that understand these languages, many are excluded from the digital revolution." However, the situation is compounded by the availability of high-quality data for AI training, particularly for less common languages and indigenous dialects, which often remain unrecorded.
While Reverie Language Technologies expands its AI translation service across Indian companies, co-founder Vivekananda Pani cautions against the potential sidelining of lesser-known dialects in favor of more dominant languages. Achieving a balance to preserve linguistic diversity while promoting tech adoption remains a challenge.
To tackle these issues, Bhattacharyya has contributed to Bhashini, a government initiative aimed at developing extensive high-quality language datasets vital for building AI models. Since its inception in 2022, Bhashini has made substantial progress, hosting 350 AI language models and facilitating over a billion tasks across different government departments.
Bhashini aims for equitable linguistic and cultural representation by creating India-specific models, steering clear of global platforms. CEO Amitabh Nag envisions a future where rural populations will access essential services in their mother tongues within a few years. "We want every citizen to feel included," he asserts.
In parallel, initiatives are underway to address healthcare challenges through AI. Associate professor Kshitij Jadhav is working to design a program tailored to assist those trying to quit smoking. This model will use empathetic communication in multiple languages to relay relevant advice to users, starting with English and Hindi.
Back in Mumbai, Vineet Sawant's efficiency has soared; he now delivers around 30 parcels per day—much attributed to his newfound language-enabled confidence through the app. "It makes us feel like we belong," he says, heralding the technology as a means of empowerment for people like him, where linguistic barriers are gradually fading in the face of innovation.
