Making a Chatbot that Teaches Languages

Alyssa Ida
Voice Tech Podcast
Published in
4 min readMar 31, 2020

There are 6,500 languages in the world.

That means that 7.8 billion people, including you and me, have 6,500 different ways to communicate. 🌏 In a world where we can go the other side of the planet in 13 hours by plane, and we’re all connected through two taps on a phone, how much of the world’s population can actually speak more than one language?

Half

That means half of 7.8 billion people cant even communicate with anyone outside of their own culture.

Not only do we need to be able to communicate with each other to survive, it is also so important in understanding the culture of other people.

But, how do we begin to tackle this? What is the most convenient way to make language learning easier?
My solution is Artificial Intelligence. More specifically, Natural Language Processing. In my most recent project, I have programmed a chatbot that supports second language acquisition through conversation.

What is a chatbot?

A chatbot is a computer simulation of human conversation. What better way to learn how to communicate than through a communication simulation?

Firstly, in order for a chatbot to simulate a conversation, it has to know what kind of conversation to simulate. You essentially have to create your own general transcript of what a potential conversation could sound like. It can be literally anything you want, this is where utility or creativity come into play.

Here’s a snippet of the dataset I used

Next, a chatbot has to process everything, hence the name natural language processing. Once a chatbot knows what to communicate, it has to break down the texts of language. It can’t just understand an entire page of text at once, it has to simplify everything into chunks of words and meanings.