India is a linguistically diverse country with hundreds of scripts, accents and pronunciations. Amidst this, English is becoming more and more important for successful careers, and yet, given the native language differences, the transition and learning from mother tongue to English is not easy. First of its kind in India, OckyPocky’s interactive and AI-powered English teaching phonetics tool aids children in recognising known and unknown sounds while enhancing their pronunciations and accents. Their huge dataset can explain English words to students through their mother tongues in Hindi and Marathi. Analytics India Magazine got in touch with Amit Agrawal, Founder and CEO, OckyPocky, to know more about the app.

AIM: What problem does OckyPocky solve? 

My roots are from Ghaziabad, with a lot of my family coming from tier three, four towns. Growing up in a small city, I realised education was a very big deal. It is the main thing you do for your child to prosper in life. Here, English is a big differentiator at every level. Later on, at my work with YouTube, I aimed to promote education and found English to be the main problem that needs to be solved. We can create all the engineers we want, but they will not have a successful career unless they fix their communication skills. The best time to learn a language is in early childhood, but most people try to solve it in their entrance exam or job interview stage. However, once your grammar or pronunciation is bad, it can not be fixed at an older age. That’s why OckyPocky aims to build a foundation for English communication skills early in childhood. Our AI-led model helps children learn this using emerging technologies.

AIM: Can you run us through the process?

We use a mother tongue approach to teach children English through NLP. I sat in 1000+ households to study how children interact with technological devices to figure out the UX required. Additionally, my grooming at Google allowed me to take a data-led approach for our second process. Data validated our gut feeling and design. This includes even small decisions as adding a lock on the homepage that may scare the users. 

AIM: What are the various tech solutions OckyPocky has built to help traditional Indian education?

We have got four products inside the app. First, there is the parent experience where parents can control, monitor and track their children’s data. This helps them gauge their progress. The second is a video corner with interactive videos that we have made unique. Third, our first-word section has won an award. It is a simple vernacular tool that teaches children how to learn English across competencies; listening, speaking and reading. Besides speech tech, we help children pronounce English properly. Finally, our offerings include first words dictionary for child building. 

AIM: Could you elaborate on the tech stack powering your platform?

Our content is vernacular heavy. Speech recognition is purely data, but everything is the data lake from a quality angle. As a start-up, your dataset keeps changing. We had the algorithm curating everything, and we have since realised children are very moody. So we had to design our algorithm around these moods. For instance, one day, the child might be interested in a red colour UI and the next day pink. And there are so many words we need to curate and teach. Our AI does this. 

To add to it, children don’t always speak perfectly; they lisp. We have to curate around their accents and their logic. Our algorithm does rule building. For instance, if there is an image of a horse and the instruction asks ‘which animal is it?’, the child might look at the colour and say brown. Our algorithm must understand the mistake and reiterate the question to make the child understand it.

Our latest addition is AI-based phonetics. We use AI phonetics to help children pronounce English properly. During our interviews with teachers, we realised the majority of schools do not cover phonetics for teachers. But worldwide, that’s the first way to teach sounds in English. Our vision was to democratise this for every parent and child. We simplify phonetics and allow a child to learn it on an app and learn the vocabulary. Our system teaches how sounds blend and how it affects pronunciations and sounds. The AI uses the same core as our first words app, where the child can speak into the app and get feedback. Our approaches include matching, reading and more. 

AIM: How do you ensure the inclusion of vernacular languages? 

Our platform has a heavy vernacular focus. We have tons of accents coming into the speech engine and needing constant refining. So far, I believe we’ve got the biggest data set in India for the age group that we deal with. This is because we are the market leaders in the early childhood space. We have thousands of users coming in every week speaking into it, and we collect this data.

Beyond your mother tongue, English is different; English in Bombay is different within the suburbs. It depends on your geographical area, academic, and social background. With each new wave of user acquisition, our sample sets keep changing. This is hard at an early set-up, but we had to keep evolving with the changes. 

Even the local languages are hard; there are so many versions of Hindi and Marathi. We are teaching English to many Marathi-English accents, and the UX helps with the transition. All our instructions are in the mother tongue. And we don’t do mechanical voiceovers; our instructions are voiced by artists that work for films and studios. We focus on keeping the instructions rich in production because children will always learn from their mother tongue. We obsess over this quality for the greatest experience. We want to curate local content, like explaining to children what is gudi padwa or vada pav or current events. We break these down to the three-year-old logic. We simplify whatever is going around the child’s life. 

We have found that many parents and children switch over to and prefer an English-English belt over a native language. 

AIM: Tell us about OckyPocky’s applications in India through case studies. 

Our users are from all parts of India; we have added our case studies on our YouTube channel from children from across the country. For instance, one of our clients was a parent from Gujarat, speaking Gujarati, and from a small town. She wanted her child to be a model and loved our assistance.