Education is crucial to childhood development, however, the distribution of resources available for childhood education is highly uneven. With over one fifth of US households spending more than 25% of their income on their children, childcare and early child education hurts a lot of household budgets. Within this picture, one important piece is children’s books. The proper development of a child’s reading ability starts at day 1 and has incredible effects on that child’s later ability to succeed in school and the workforce.
This importance makes it no surprise that the US children’s book publishing industry is big business, to the tune of $166 million profit on $2.3 billion of revenue annually. Within this, ebooks make up about 12% of the industry, or $244 million. Within this industry there are unaddressed problems. A study conducted earlier this year found that there are all too often biases in children’s books. Female characters are grossly underrepresented in children’s books. And when they are, they tend to be the sidekick.
Machine learning technology has advanced to the point that it is very close to creating believable human prose. However, it still falls short given the complexities of adult language and is unable to create plots that will outdo human creativity and logic. Children’s books, however, offer a potential area where machine learning could apply. Not only is the syntax and structure of child language far simpler than adult language, but the plot does not need to hold up to the same level of logical scrutiny and complexity. By ingesting large volumes of children’s books and utilizing some of the cutting edge techniques of data science and natural language processing, we believe it is possible to algorithmically create children’s stories.
AI of Solution
The way we could build this system would be to incorporate several different machine learning tools to generate and evaluate stories until a threshold of acceptability was reached. We would ingest a large volume of data, likely from open source books to begin with. From there, we use tools such as word2vec, sentiment analysis, and topic mapping to generate stories, and use guidelines from the latest research into child development regarding sentence structure and vocabulary to ensure simplicity and age-appropriateness. Using massive training sets, a recurring neural network (RNN) will use layered outputs that are then reused in the inputs. Cautious of overfitting, one tuning parameter is the number of epochs the data is used. For the shakespeare code it was 30. Another challenge of the model is the short-term memory. The algorithm cannot remember “long-term” and so an architecture being explored now is LSTM and GRU, using gates within the code. Adding additional layers of gates could help to find higher-level interactions but the more layers we choose, the more training data we need.
The long term vision for this product would be an application that allows the user to feed it the reader’s age, some topic to begin with, and a profile of a character they would like to see. This way, underrepresented groups could find themselves the protagonist in a story of their own creation. In addition, using technology similar to that used in real-time digital advertisement generation, we could programmatically include pictures to follow along with the story. If the technology is strong enough, we would hope to build partnerships with various children’s entertainment content providers. Kids all have their favorite characters. We want to let them generate their own stories with those characters too.
In order to pilot this concept, we would first need to ingest a lot of data. Luckily, there is a wealth of children’s books available within Project Gutenberg. Less luckily, these are mostly written in the 19th century. That means that they contain outdated language and social ideas that are largely shunned in most modern children’s stories. However, for a proof of concept, this is acceptable.
Once we have data, we can create a genetic algorithm to generate simple sentences using the book data we ingested, word mapping algorithms, and a list of age appropriate vocabulary to create sentence. We can evaluate these in two ways. Using sentiment analysis, we can evaluate based on how well the sentiment of the story follows the sentiment of training data stories. And using topic mapping and word2vec, we can push the stories to transition topics gradually and make logical sense, rather than jumping around incoherently. Using the generation and scoring algorithms, we can use a genetic algorithm to optimize both the story arc and the story coherence. From there, we can display it to a young reader.
We believe that our app is commercially viable given our initial research and interviews with parents with young children. Based on the feedback from the data scientists from the last class, our research seems promising and is something that can be executed, especially given the fact that children’s stories are simpler than complex stories or writing for adults. Further, we have created a survey for parents through which we hope to further demonstrate the commercial viability of our app.
Interview with parent Nitin M. (student at Booth, parent of 3 year child): They buy 3 books a month worth around $40 books, so about $500 in the year. Their only concern is that they prefer giving their son physical books versus tablets. His concern about tablets is reducing though considering that the technology with screens etc. has been improving vastly. The quality of stories is very important to him and he doesn’t want the quality to deteriorate. “If the AI app can provide good quality throughout it would make more sense to continue. It shouldn’t be like Netflix where after you watch 10% of the content the remaining 90% is not great.” Additionally, some suggestions he had were: “If there is a picture form and if you can click on a character if the character does some type of action. For example, his kid loves Aladdin, so it would be cool if within the story he could click on a sentence and see Aladdin perform some action.”
Interview with Adi (International student at Booth, parent of 1.5 year old child): He has thought about something like this specifically because he’s from India and the books he buys in the U.S. are more focused on American characters with American names. According to him he would love this sort of a product because he would be able to have customized stories with Indian names and characters that his child does not get to experience as much while living in Chicago. His concern with the product is that if he has to input too many things it might become tedious to use our app and he might stop using it after a while. Otherwise he is sold on the concept and he likes that the app would evolve stories for children as they grow older with age. If the app can plug in to the parents life and get some indicators and utilize those to create stories it would be easier for him to use. We should aim to provide an easy “default case” that can readily be used for his child.
We also analyzed our competition: Epic! Which is a digital library for kids with a library of over 25,000 books both for kids and for educators. The price is $7.99/month and it has over 44,000 reviews on the app store. While they have an early mover’s advantage they do not have the option to customize based on various factors such as gender, race, citizenship of the parents etc. which will be our unique selling point and provides us with an advantage. We also plan on having a freemium model with up to 3 free stories in a month which will be based on a basic initial input into the app. For anything more interactive or customized, we plan on charging $8/month (similar to our competition) to our users as an upgrade cost. Based on the fact that the children’s story book market in the U.S. is worth $2.3 billion we have a huge potential even if we target the users who only use e-books.
Costs: Based on our research and from discussing with the data scientists in class we assume that we will need one programmer full-time ($100,000 a year) plus two part-time programmers initially to get the application setup. We will conduct a hackathon within the Chicago tech community to build a piece of the app out with data that we get on our own and consider hiring from among the best programmers from that hackathon.
Link to survey sent out to parents: