(Subscribe to our Today's Cache newsletter for a quick snapshot of top 5 tech stories. Click here to subscribe for free.)
The dream of building a startup took Sourabh Gupta and Akshay Deshraj, IIT-Roorkee alumni, to Bengaluru. To understand the next billion customers and their needs, they travelled from village to village across, and their search came to an end when they met a farmer who told them about the difficulties he faces while engaging with the bank. The farmer explained that the messages he receives are in English, a language he does not understand, so he has to travel 10 kilometres to get that translated and in the process he loses out on his day’s earnings.
Around this time, Sourabh was visiting his hometown in Punjab. He alighted at Chandigarh airport, collected cash from an ATM, and left for his home in Ludhiana. On reaching, he realised that he had left the debit card at the ATM.
As the drive to the ATM was two hours, Sourabh decided to block his card through the app. But he could not figure out how to. He tried the same with Internet banking and failed again. Eventually, he was able to block his card through a call centre which took him almost 30 minutes including three attempts at finding the right key sequence to reach the call centre agent.
- One challenge with speech is understanding emotions of the users. Sourabh points out that one can unlock significant value once the machine figures out whether the user is angry, happy or sad. That is one of the primary research areas Vernacular.ai is working on.
- Sourabh believes that certain things, such as reporting fraud, should not be handled by a bot as when a user is agitated and frustrated, a human being should come in to assist to keep some processes holistics
The multilingual barrier for rural banking users, coupled with the communication issues on call centres, led to Vernacular.ai. “Something as simple as that should not take that much time,” Sourabh, co-founder and CEO of Vernacular.ai, points out over a phone call with MetroPlus .
The Bengaluru startup is an Artificial Intelligence-first SaaS (Software as a Service) business that is enhancing customer experience through voice conversations. It provides a multilingual AI-based automation platform named Vernacular Intelligent Virtual Assistant (VIVA), designed to engage in natural conversations.
The company helps enterprise customer contact centres automate customer queries using VIVA. The assistant has the ability to understand over 16 languages and more than 160 dialects. The platform can help automate up to 80% of call centre operations.
How it works
The importance of technologies such as AI and voice technology became more pronounced during the pandemic. As travelling was limited, more and more people shifted to digital methods of engagement.
Sourabh believes that customers want quick query resolution and easy access, and multilingualism is a critical aspect of this since more native-language speakers are coming online. So translation, transliteration, and speech in regional languages are important in today’s landscape.
“A couple of hundred million Internet users in India are not availing digital banking because such facilities are not available in their regional languages,” Sourabh explains. While people do not understand English, many do not understand multilingual keyboards as well, he adds.
This is why voice is going to be a very intuitive way to engage with interfaces online. Vernacular.ai aims to help regional language speakers communicate with call centres in their languages and get solution to their queries.
The technology and algorithm built by the Vernacular.ai team has two critical engines: speech-to-text and text-to-speech.
The complexity of dialects, accents and languages in the country made it difficult to train the systems. Hence, data was collected from every region separately, so the machine could understand and differentiate between different dialects and accents. For instance, there is difference in how a person from Bihar would speak Hindi and how a Delhiite would, and the system was trained to distinguish between the two.
Sourabh points out their machine is trained not only in languages and dialects but in local and Indian cultural nuances as well. As algorithms are specifically built for these languages, accuracy of Vernacular.ai is 15 to 20% higher as compared with any other speech engine, he added.
‘A world of voice command’
Talking on the future of AI and Voice, he says, “We are heading towards a world of voice command where many interfaces will cease to exist. AI will add a trillion dollars to the Indian economy by 2035 and the market is going to explode in the next few years.”
Sourabh says that two years ago, they had only one customer but today they serve more than 25 enterprises in multiple sectors such as Axis Bank within finance and Barbeque Nation in food and beverage.
He adds, “We have scaled out, expanding to Southeast Asia and the US. Expanding to these territories made sense because Southeast Asia is similar to India in terms of language complexities and nature of customers. US is the most mature market for our technology and becoming a global leader would not be possible without entering US.”