At a loss for words?

When Prime Minister Narendra Modi met digital content creator Shraddha Jain recently, he greeted her with a peculiar South Indian exclamation – 'Aiyyo' – that, in certain contexts, would be considered inauspicious. Modi, however, was not being ungracious: Jain has built a reputation for her comedic reels on social media platforms under the brand identity of Aiyyo Shraddha.

The word 'aiyyo', which made its way into the Oxford English Dictionary in 2016, is laden with linguistic latitude. As the OED notes, it can be invoked to convey a range of emotions and responses: surprise, disbelief, disappointment, exasperation, regret, distress, grief (and, counter-intuitively, even joy), and much more. It is even synonymous with the ejaculations "Oops" or "Oh My God". In most cases, it's the auditory tone that conveys the precise meaning. For that reason, a rudimentary understanding of cultural contexts may be necessary to avert miscommunication of intent when invoking that versatile word. Also for that reason, and particularly given India's linguistic diversity, words like this pose peculiar challenges in an unlikely domain: the creation of chatbots and language models powered by Artificial Intelligence (AI).

ChatGPT, the AI chatbot, has impressed many, but it struggles to respond in Indic languages. In India, where 22 languages are constitutionally recognised, achieving language parity with English in AI technology is an important goal. That is the mission of a network of engineers and researchers currently creating machine learning systems for Indic languages.

This is particularly important since a wide swathe of the country uses Indic material to access Internet services. Over 205 million potential users cited Indic languages as a crucial factor in their decision to use the internet, according to an estimation by the Internet and Mobile Association of India (IAMAI) and consulting company Kantar (see box 'Lingo bingo'). They forecast that there would be 900 million internet users in India by the year 2025, with voice and Indic languages serving as important growth drivers.