Exploration of Temporal Knowledge Graph for Story Analysis

SynapTale has developed a temporal knowledge graph model that represents a story through nodes (entities) and edges (actions and relationships). The current demo features 232 entities and 1,852 edges, reflecting the story's state at each chapter. Notable findings include the character with the highest kill count being the Tin Woodman, and Dorothy not deceiving anyone in the first 100 chapters. The model employs various multi-agent pipelines that integrate large language models (LLMs) and natural language processing (NLP) to construct the graph and validate relationships over the chapters. The system also introduces epistemic nodes to capture different perspectives on facts within the narrative.