Sibil AI: Children Story Generator in Sinhala using Transformers

Abstract

Ever since the birth of humankind, stories have been used as a means of sharing information and educating people. Stories are more than just a form of entertainment; they impart lessons that often help children to develop the skills they need to thrive in life. Research and development teams have unquestionably mastered the practicalities of producing human-like creative text tales, which has been a significant barrier in natural language processing in recent years. A system based on artificial intelligence that generates children’s stories can serve as a resource for parents and children to connect with. The ability to generate natural language stories that people can understand, remember, and enjoy is difficult to achieve with current technology. A new model based on transformers is introduced in this paper. This new approach for generating stories for children based on the GPT-2 model with the help of a web application. The GPT-2 is a model based on neural network that is designed to imitate the human behaviour of producing creative and coherent text. It can generate stories in different genres and starting captions. The web application takes advantage of the GPT-2 model’s ability to generate fluent texts, including proper punctuations, complex syntaxes, and grammar rules. The solution allows users to generate creative stories from different genres with starting captions. Especially, using the proof-ofconcept to support the narration given in Sinhala language, one of the native languages in Sri Lanka.