Boom time for Indian-language virtual assistants

India, the second most populous country in the world and home to over 20 prominent languages, is on the threshold of becoming a driving force in the world of voice-enabled virtual assistants.

A mini explosion of start-ups and investors are placing their bets on vernacular-powered virtual assistants that will cater to an Indian market of over half a billion people, as the artificial intelligence (AI) sector eyes mobile users who are more comfortable working in Indian languages than in English.

Making waves regionally

Reverie Language Technologies, a Bangalore-based start-up founded in 2010, launched its own Indic voice suite, Gopal, last year. Gopal is powered by an automated speech recognition engine in 11 Indian languages and can power voice bots and interactive voice response (IVR) to boost user engagement.

“Most Indians do not speak English. This is a fact. Most of the Indian Internet is only available in English. This is also a fact. What this means is that most Indians effectively have very little space for themselves online,” says Arvind Pani, co-founder and CEO of Reverie Language Technologies.

Kavita Reddi, co-founder and director of Voxta Communications, a speech recognition start-up founded in May 2014, says, “It is like Alexa or Siri in Indian languages.”

Voxta Communications provides bespoke voice bots across three sectors: banking, e-commerce and education. Services in English and local Indian languages like Hindi and Telugu are currently available while voice services in Marathi, Gujarati, Kannada, Bengali, and Tamil are in the pipeline, confirms Reddi.

Rapid growth

Reddi believes that the widespread availability of smartphones makes for an entirely new potential online audience. People in this sector may not necessarily be literate or fluent in English, yet using smartphones to access information online by typing in Hindi, Bengali, or Tamil, she adds.

Deloitte published a report last May, predicting a growth in technology, media and telecommunications. It stated that India had nearly 432 million internet users, and that this number is growing rapidly. They expect that vernacular users in India will be more than 2.5 times the English internet user base by 2021.

These findings are backed by a KPMG report released last year, which asserts that India’s Hindi internet user base, along with vernacular languages like Marathi and Bengali, is expected to grow. It further states that, up until 2021, users of South Indian languages will be among the most digitally engaged in the country. This is because Indian language users will AI assistance in their own languages more reliable than English content. These users are likely to account for almost 75% of India’s internet user base by 2021.

Investments, too, are on the rise. Rijul Jain, associate principal at Astarc Ventures, acknowledges the need to produce content in vernacular languages. “As long as this need is not satiated, the number of start-ups and investments in this sphere will keep growing,” he says. Astarc Ventures has already invested in AI-driven vernacular start-ups, namely Liv.ai, Khidki, Norah.ai, and Infinite Analytics.

Microsoft Accelerator has put money into Bangalore-based Reverie and Aspada, while international investors include Qualcomm Ventures, a California-based investment firm. Niki.ai, another start-up, has raised seed funding by Ratan Tata and Ronnie Screwvala-owned Unilazer Ventures.

Highs and roadblocks

The use of AI is a sophisticated process and its success depends on how accurately it can deliver services.

Sachin Jaiswal, CEO of Niki.ai, says AI, coupled with natural language processing (NLP) and machine learning, is helping them understand the way people converse and give accurate responses.

“With the language-agnostic NLP engine, backed by data of over three years now, Niki can understand multilingual inputs in a single sentence, removing all language barriers and enabling smooth flow of conversations,” he says.

Niki provides a chat-bot service that allows users to talk to or text Niki to shop for numerous products and services, like booking movie and bus tickets, hotel and cab reservations, making utility bill payments, carrying out mobile recharges and laundry services.

However, start-ups are not without challenges. A PwC study conducted in 2018 to understand the impact of AI on Indian businesses found several deterrents to growth. The key barriers included high costs, lack of technical ability and quality data and privacy concerns.

Jain believes that for AI to function well, a high volume of data is needed, and the challenges for start-ups depend on their individual goals.

“If someone is working with a voice product that caters to a Hindi-speaking user, you would need maximum recordings in Hindi to train the model. Here, the challenge could be the kind of data in terms of variation, language and dialects,” he said.

Gupta, on the other hand, says some of the world’s biggest technology companies have already shown how to use big data, and this can be used in both a positive and as well as a negative manner.

“Leading technology firms have taken the big leap and are steadily gathering customer data and understanding patterns of people’s consumption based on their preferences,” he added.

“These companies do (experience) litigation with regards to data privacy, but in India the guidelines are not too clear. There is a very thin line between what is acceptable and what is not.”

Government involvement

With flagship programs like the Make in India, Digital India, and Startup India, the government seems to have embraced the digital revolution. This has also highlighted the need in India for an efficient Data Protection Bill, which is currently at a draft proposal stage.

The government has also worked closely with start-ups in the sector. In 2014, Voxta Communications tied up with the Bharatiya Janata Party (BJP) and launched the IVR campaign where users could dial in and listen to Prime Minister Narendra Modi’s pre-recorded messages or even live speeches.

Reverie, too, has worked closely with the government on its e-Marketplace, where they helped to create the platform in Indian languages, removing English as a barrier to usage for ease of adoption by all government officials. The company has also cooperated on the e-National Agriculture Market, popularly known as e-NAM, which has enabled real-time access to pricing data and a marketplace for millions of farmers in their own languages, cutting out the middlemen.

Voxta has its own English learning app — Chat Chit — that teaches people spoken English. “Voice is the critical bridge, and its moment has come,” Reddi says.