Automated Story Conversion: Generating Children stories from Adult stories

1. ABSTRACT We developed an approach to convert“adult stories” to“children stories.” Specifically, we obtained 30 stories, 15 of which were of each story type. Our Lexical Analyzer (LA) made use of OpenNLP’s part-of-speech tagging utility. With this, we developed a means to do clausal separation, allowing our story to be more direct and explicit by generating more simple sentences. We looked at a metric we called AdultLikelihood in order to determine which words are the most technical and adult-like. Our aim was to target these words and replace them with simpler, child-like words. Our system implemented WordNet’s Synsets in order to find these candidates, and the actual replacement words will be picked based on their Adult-Likelihood value. We additionally included OpenTextSummarizer into our project, to allow the user to optionally trim out sentences that have rarely occurring words. Regardless of that added feature, we then evaluate our performance by measuring “Sentence-Complexity” in various ways. Lastly, we developed a neural network that classifies stories as being of type adult or children, based on key features that we identified.