Generate Causal Story Plots by Monte Carlo Tree Search Based on Common Sense Ontology

To automate the generation of a story, the computer systems require sufficient amount of causal knowledge and need to synthesize a feasible plot from a larges scale of possible causal sequences. This paper describes a Monte Carlo Tree Search (MCTS) method to find a sequence of causal events as a story plot from a large scale of common sense causal ontology. A common sense causal knowledge base is established first by extracting various causal links embedded in ConceptNet5 and by classifying the concepts in the causal links according to the Fabula elements. We show how MCTS finds a feasible sequence of causal events as a story plot according to the Fabula model given arbitrary initial and goal concepts and a desired story length.