Event similarity based on the analogical mapping between contextual structures in stories: A computational model

Abstract

In a story, the meaning of an event depends on its context or its relationship with other events. To elucidate the cognitive mechanism associated with the construction of such contextual meaning is a fundamental issue from the perspective of cognitive science and artificial intelligence. Hence, this study proposes a computational model that determines the similarity between two events occurring in different stories based on the analogical mapping between the events' contextual structures. The proposed model automatically constructs the contextual structure of a chosen event by associating relevant verbs gathered from precedent and subsequent events. Similarity between the selected events occurring in two stories is thereby estimated on the basis of the mapping between their contextual structures. To assess the proposed model, it was compared with human ratings of event similarity by incorporating twelve pairs of experimental stories. The obtained result indicates that a moderate or a strong positive correlation can be observed between the proposed model and the human ratings.