Narrative knowledge graph

Definition

A narrative knowledge graph models the story as a network: nodes are entities (characters, places, objects, events, world concepts), edges are typed relationships ("is allied with", "is located at", "owns", "knows that", "precedes chronologically"). Unlike a linear prose bible, the graph supports structured queries: "which characters had access to information X before scene Y?", "which places connect to character Z?"

It is largely fed by Narrative entity extraction from the manuscript, then consolidated in the project's AI contextual memory. The graph does not replace literary text: it is a queryable map for a Narrative copilot (AI) or Automated consistency analysis. A temporal graph — where relationships evolve scene by scene — separates what is true in chapter 5 from what held in chapter 1.

What it's for

The graph makes Internal coherence checkable at full-manuscript scale. In mystery fiction, it maps who could have been present at the crime scene; in fantasy, it traces alliances and oaths; in romance, it tracks how emotional bonds evolve. For the author, it is a structural mirror: sometimes the graph reveals an orphan character (no useful relations), a faction never mobilized, or a planted object never recalled.

On the AI side, the graph supplies compact, precise context: instead of injecting 200 pages into a prompt, you query nodes relevant to the current scene. That reduces hallucinations and improves copilot suggestion quality. The graph also documents authorial intent separately from draft text — useful when a planned alliance exists in notes but has not yet appeared on the page.

Variants by genre

• Fantasy / SF: multi-typed graph — factions, magic systems, artifacts, lineages; power and technological dependency relations.
• Mystery / thriller: temporal and epistemic graph — each fact timestamped, each "knows that" relation dated; essential for alibis and late reveals.
• Romance: relational graph — tensions, breakups, reconciliations; edges carry emotional attributes, not only facts.
• Literary fiction: lightweight graph — focus on core characters and motifs; avoid over-modeling a still-fluid draft.
• Series: versioned-by-volume graph — merge or split nodes across books, manage explicit retcons.

Common pitfalls

• Graph built, never used: upfront investment with no integration into writing or revision workflow.
• Over-modeling: everything becomes a typed relation; the graph becomes unreadable and costly to maintain.
• Abandoned manual updates: the manuscript evolves, the graph stays frozen — worse than no graph.
• Confusing graph with prose bible: duplicating sheets without using relational queries.
• Noisy extraction: without reliable Narrative entity extraction, the graph accumulates false positives and unresolved homonyms.
• Static graph for unreliable-narrator fiction: "knows that" relations must reflect character knowledge, not world truth.

Examples

• Les Misérables — filiations, oaths, and encounters form a dense network; a graph would visualize Valjean, Javert, and the Thénardiers' paths across volumes.
• And Then There Were None — a temporal graph of island presences and successive deaths is the natural tool to verify puzzle logic before publication.
• The Lord of the Rings — alliances, rings, elven and human lineages: a relational graph is ideal support for Automated consistency analysis at that scale.

Writing tips

• Feed the graph as you write, not in post: rebuilding a graph over 100k words takes 20 hours; feeding it live takes 5 min/chapter.
• Distinguish "fact" nodes from "rumour" nodes: the reader thinks X but the truth is Y. A graph without that lies to future rereaders.
• Query the graph before each new chapter: "who knows what at this point?" prevents 90% of info-knowledge inconsistencies.
• Archive one graph version per book at end of drafting: 2 years later you won't reconstruct the exact end-of-book-1 knowledge state.

Learn more: Narrative threads: stop forgetting your story's promises and Why the Relationship Graph is the Secret Weapon for Complex Novel Authors.

Revision checklist

• Do AI contextual memory and this notion reinforce each other, or does one contradict the other?
• Could a beta reader name Narrative knowledge graph after reading without being told?
• Am I keeping a scene out of attachment that doesn't illustrate this notion?
• Did my last pass fix structure before style (see Line editing / polishing if needed)?
• Does the climax or pivot scene clearly show what Narrative knowledge graph changes in the story?