Generating children’s stories from character and event models

• Main Authors: Adolfo, Bianca Trish, Lao, Jerson, Rivera, Joanna Pauline, Talens, John Zem, Ong, Ethel Chua Joy
• Format: text
• Published: Animo Repository 2017
• Subjects: Natural language generation (Computer science); Computer fiction; Knowledge representation (Information theory); Natural language processing (Computer science); Computer Sciences
• Online Access: https://animorepository.dlsu.edu.ph/faculty_research/450; https://animorepository.dlsu.edu.ph/context/faculty_research/article/1449/type/native/viewcontent
• Institution: De La Salle University

© Springer International Publishing AG 2017. Stories are inhabited by multiple characters who interact in the story world to achieve their goals. For computer systems to generate such stories, we propose the use of agents to model the needs and behavior, as well as formulate the individual plan of action of the different story characters. A central agent, the story planner, uses a simple turn-taking approach to maintain control over the asymmetric roles and goals of the character agents. It coordinates the various events that will be allowed to take place in the story by prioritizing candidate actions from the character agents based on their likelihood of increasing interaction, which may be verbal (conversations) or nonverbal (helping one another, playing together). The resulting stories are perceived by the human evaluators to have good character-to-character interaction, but acceptable level of coherence and cohesion of events in the story text.