Neural Text Generation Model for Personalized Storytelling in Primary Education
Abstract
Individual narration in primary schools is crucial in developing literacy, creativity, and interest among the young learners. The process of adapting the stories to interests and reading levels of a specific reader promotes understanding and motivation during early learning stages. Current digital storytelling techniques are mostly based on generic content or text generation rules that are not personalized and fail to meet the needs of different students and, thus, yield less engagement and less optimal learning. In order to overcome these shortcomings, the given paper suggests a Neural Text Generation Model based on the conditional variational autoencoders (CVAE) to create personalized stories. The CVAE framework can be used to produce contextually adequate and age-appropriate narratives that are conditioned on the profile of the students, such as their reading skills, interests, and learning goals. The suggested model can give both the teachers and the learners dynamically generated story content thus creating interactive and customized learning experiences. Evidence of experiments shows that the model has a great impact on increasing the engagement, comprehension, and motivation of students, when compared to the traditional storytelling methods.