At YourStory’s Future of Work conference, Janu Verma, Senior Data Scientist at Hike, reveals why Hike is focusing on stickers and how the ‘upgrade on emojis’ is changing the way Indians communicate.
What if you could replace words with stickers on your messaging app? Hike, India’s first homegrown messaging service, wants you to do just that.
At the second edition of YourStory’s Future of Work conference, a first-of-its-kind gathering of technologists, developers, data scientists, researchers, hiring managers, entrepreneurs , and corporate leaders, Janu Verma, Senior Data Scientist at Hike, spoke about what lies beyond the QWERTY keyboard and how Hike might just be able to replace the old keyboard with its stickers.
Hike is the first experience of working at a product company for Janu, who has a background in pure mathematics and physics. He was a former Research Engineer at IBM and former Quantitative Researcher at Buckler Lab for Genetic Diversity, Cornell University.
We may be in the 21st century, but the QWERTY keyboard, invented in the 1870s by American inventor Christopher Latham Sholes, remains the primary input method.
“It is really shocking that it continues to be the mode of input in India even when people try to communicate in a local language,” Janu said.
He added that India has nearly 500 million internet users today and majority of them spend their time on some kind of messaging application, be it WhatsApp, Facebook Messenger, or Hike.
“Despite spending so much time on messaging apps, we Indians are crippled by the outdated keyboard system we use,” Janu said.
While countries like US and China have one major language, India, with more than 20 languages and scripts, has only one keyboard. While some work has been done in this direction with the invention of Google's Indic keyboards, these local language keyboards are still complex, cumbersome to use, and not mature enough. They do not offer an auto-correct or quick-typing facility. Moreover, there is a language problem as people spell different words differently.
“People are more comfortable with the Latin keyboard. Even when I type in Tamil, I feel more comfortable using the Latin keyboard,” Janu said.
This is the problem that Hike’s stickers– one of the platforms most loved features - are trying to solve. Hike offers users a library of over 20,000 stickers in more than 40 languages, covering multiple genres that showcase the colour and culture of India.
The entry of voice and video messages has led to a major shift in the way we communicate. However, there is a massive UI problem. “Think about attending to a voice or video message in the middle of a meeting,” Janu said.
He said Hike is working to solve this problem with stickers. “Stickers are visual expressions, an upgrade on emojis,” he said, adding that the advantages of using stickers as a medium of communication were many. Team Hike has taken great care to “Indianise” their stickers, which come in a wide range from Bollywood to pop culture.
Most Hike conversations comprise short messages and go on for hours. To make sure that these conversations take place through stickers, Hike has to “have stickers for every potential conversation that we are trying to have”.
Hike messenger has observed that people use up to 342 variants for the phrase 'Kya kar raha hai?' . The company wants to expand their sticker bank 10x. But how does Hike ensure there’s a sticker for everything you want to say? Hike has a set library. The different features and components - faces, eyes, lips, skin tones, and background colours - are assembled to make the different stickers.
“The stickers have to be aesthetic and relevant. We cannot have a Rajasthani sticker saying something in Marathi,” Janu said.
Knowing which text should go with which sticker is also a task. However, having a large bank of stickers may also be a problem. “It is a terrible user experience if we make our customers go through all the stickers to find the one they are looking for,” Janu said.
To solve this, "We (Hike) have to make the discovery pipeline 10 times better than what we have today,” Janu said.
Hike is developing a sticker recommendation model or a sticker discovery pipeline, which will make it easier for the user to pick the sticker up, he said.
Hike is also trying to capture the language used in the conversations.
“We have a neural-network model, which - from the context of the conversation - tries to figure out the possible response (sticker). As you are typing, it tries to figure out what is it that you are about to say,” Janu said.
User information, mainly age and gender, and other inputs takes the predicted chat that the user is about to write and maps it to the possible sticker for the conversation. Hike also has a model that identifies similar stickers or stickers with similar phrases. Finally, Hike is trying to figure out the chat context; learning which stickers are mostly used in which situations and at what time of the day.
“We are trying to understand the context of the conversation whether an user is conversing with their a friend, mother, or girlfriend. We are trying to figure out the possible relationship people could have,” Janu said.