Show HN: Oz Series as 180k-Word Temporal Knowledge Graph

SynapTale offers a free, registration‑free interface that lets users explore a dynamic temporal graph of a story. The system constructs a network of entities (nodes) and their actions or relationships (edges) rather than simply visualizing a wiki. Each chapter is captured as a snapshot, allowing the graph to retain details such as a promise made in chapter 8 even after 100 chapters. The current demo contains 232 entities and 1,852 edges, with analytics and timelines derived from the underlying graph.

Analysis of the first 100 chapters reveals several notable patterns: the Tin Woodman has the highest kill count, followed by Dorothy with three killings; Dorothy never deceives anyone in this period; the Scarecrow’s debt to the stork has remained active for 92 chapters; the Cowardly Lion ranks third in threats; and the chapters include 60 secrets and 254 dialogue events. Technically, the platform employs five multi‑agent pipelines that combine large language models and natural language processing for prescanning, ontology construction, chapter‑by‑chapter graph extraction, retrospective validation, and linguistic profiling. To manage time‑dependent relationships, the system defines three edge types—event, identity, and state—ensuring that most edges terminate within the same chapter, which improves scalability. An ontology scan precedes extraction to standardize entity and edge types across the narrative, and epistemic nodes are introduced to capture differing perspectives on the same fact.

The SynapTale demo demonstrates how a living story can be represented as a living graph that evolves with the narrative. The platform’s open‑access interface and detailed analytics provide researchers and enthusiasts with a new way to examine story structure and character dynamics. The demo is available at https://news.ycombinator.com/item?id=49053986.