Artificial intelligence in storytelling



Da Hyeon Choi

Posted: 7 November, 2018

Da Hyeon Choi, winner of the Irish Research Council Award for Excellence in Interdisciplinary Research at the EU Contest for Young Scientists, tells us about her work on storytelling and artificial intelligence for the #LoveIrishResearch blog. Da Hyeon is currently in 11th grade at  Cheong Shim International Academy in South Korea. She is interested in technology, literature and sociology. The EU Contest for Young Scientists was set up in 1989 to promote the ideals of cooperation and interchange between young scientists. 

She began as a simple concept when I was thirteen. I’d been teaching my younger brother English, and the countless nights of reading him bedtime stories had reminded me of my own childhood, with my mother at my bedside. At first, I wanted to build a smart bedtime story-teller modelled after this experience, a program that would essentially be a babysitting helper.

The actual development began a couple years ago, when I began teaching at the local library and orphanage. The idea that I’d been fostering since I was little became a full blown project, as I slowly realised the implications (and possibilities!) of an artificial intelligence like LYRA (a Learning Youth’s Reading Assistant). Then as I got more invested, and started looking into the significance of storytelling.

Have you ever wondered about why storytelling is so common in various cultures? Why it might be important? As it turns out, storytelling is crucial for the development of literacy in children, as well as the creation of their idea of self.

The problem, I realised, was that a lot of children today don’t get an opportunity to experience storytelling. Busy parents, absent parents, illiterate parents… there are reasons for this, but the best way to solve the problem, I thought, would to build an easily accessible storyteller AI.

See, LYRA isn’t a simple audio-book. What she does — and what she will do in the future — is hold conversations with children according to the context of the story she’s reading to them. For this purpose, her prototype (the one I introduced at EUCYS 2018) relies on three main parts: her speech synthesis and recognition software, her deep-learning system based on story-centred data, and her dynamic narrative. The prototype creation process involved a lot of research, frantic chats, coding errors, and late-night caffeine. I had a lot to learn and a lot to test out.

The data of LYRA’s current prototype is based on Snow White. Part of my project involved testing how children would actually feel about talking to a program, and how engaging LYRA could be. Thankfully, most of the children loved the experience!

At EUCYS, I was able to test her out with strangers, and an especially interested group of students gave me ideas on how to make LYRA more child-friendly. All of the advice I got in Ireland was both enlightening and encouraging, and, I have to say, it was an invaluable experience that really fuelled my determination.

It wasn’t an easy process, building LYRA and planning out ways to introduce her concept to people. She still has a long way to go, and I hope the world gets to see her in action soon.

Disclaimer: The opinions expressed in our guest blogs are the author’s own, and do not reflect the opinions of the Irish Research Council or any employee thereof.