History has witnessed many instances where the darkest hours have inspired the brightest ideas out of people. Thirty-four-year-old Aakrit Vaish, CEO and co-founder of Haptik, an enterprise conversational Al platform, would surely agree to this. After all, his WhatsApp chatbot—MyGov Corona Helpdesk— is helping millions in the country to address queries around the novel coronavirus outbreak which in a way has brought the world to a standstill. The chatbot powered by AI has come at a time when misinformation regarding the virus is spreading like wildfire.  

“We understand the severity of the situation and the idea for the Coronavirus chatbot germinated as a medium to tackle misinformation,” says Vaish. “The ‘MyGov Corona Helpdesk’ has been engineered to fight rumours, educate the masses and bring a sense of calm to the current chaos-like situation. We are committed to assisting the government with all our possible strengths and resources and hope that this chatbot can help the GoI spread the right information across the nation.”

The WhatsApp chatbot has been a success, to say the least. As of April 1, it has processed more than 27 million conversations with users ranging from across the country. “Moreover, more than 18 million users have taken their queries to the chatbot,” says Vaish. As expected, the most common queries has been on coronavirus symptoms and on how one can protect themselves. Other queries like ‘how does the coronavirus spreads’ is also common. 

“When my co-founder and I started Haptik in 2013, we did so out of a belief that conversations are the most native interface for the smartphone in the 21st century. We’ve always been convinced that this technology has wider social applications. It is our hope that this endeavour, in addition to raising public awareness in a time of crisis, also heralds a greater role for Conversational AI in all spheres of life,” he says. 

Essential to fight rumours

People can access the helpdesk chatbot on WhatsApp by messaging on the number- 9013151515 and get access to information related to the COVID-19 pandemic in a conversational manner. “This includes information about the coronavirus, its symptoms, and how it is transmitted. It also gives information on the various preventive measures that can be taken as a safeguard against the virus; health & travel advisories; and official government helplines for further assistance,” says Vaish. 

The Helpdesk was launched on February 10 much before the virus spread its wings in India. However, it was only in March that the team began building the chatbot in collaboration with the government. “In March, we began working with the Government of India to build a chatbot which was similar to our previous chatbot,” says Vaish. Having said that, the ‘MyGov Corona Helpdesk’ is a completely new chatbot that has been built from the ground up in four days, he clarifies. 

People can access necessary precautionary measures, symptoms, correct and incorrect facts, helpline numbers, affected cases in the region, government advisories (including travel), informational videos, etc.

Leveraging AI

The ‘MyGov Corona Helpdesk’ operates within a boundary of defined questions and answers. The data captured from interactions are being used to improve the chatbot says Swapan Rajdev, co-founder and CTO of Haptik. He shares details on how the bots are being trained using AI. 

“The fundamental concept behind a bot is that whenever a user comes in and types anything, it [bot] should be able to understand what the user is saying and formulate a response to it,” he says. He explains further by an example. “Suppose there is a user who asks directly about symptoms of corona [COVID-19]. There is another user who asks if fever and cough could mean having corona, then essentially the answer to both the queries is the same. So a bot should be able to handle this,” Rajdev says.

In order to train the bots to handle such varied queries, there is a lot of machine learning that is involved along with natural language understanding, which is also part of AI. “We have developed some conversational flows to understand what queries users can potentially have. This is an important element of bot building. We get data based on this conversational flows and accordingly train our ML models,” he says. Whenever a user asks something, the bot’s goal is to follow or refer to these conversational flows.

Swapan Rajdev and Aakrit Vaish

The problem areas

Once the bot went live, Rajdev faced a couple of problems—common for most bots. “Sometimes, it does not understand what the user is saying. Moreover, bots trained by AI can have a problem of false positive,” he says. 

“Apart from constant training, we have a manual layer of vetting in the middle, and we call it a semi-automated feedback mode. Since coronavirus is a sensitive topic, we do not want any wrong information going to the user as a false positive. So every easy single query is vetted by AI. The ones that cannot be understood by the bot gets looked into by humans.”  

The spread of fake news

Ever since the coronavirus outbreak, there have been multiple instances of fake news or misinformation getting circulated via WhatsApp, which has more than 400 million users in India. Prime Minister Narendra Modi also appealed to various organisations to fight against misinformation about coronavirus. 

“We hope the initiative plays a big role in our fight against the virus as well as fake news and misinformation,” Vaish signs off.

Image Source: Pixabay