Skip to content

Knowledge Graph Framework -- Documentation

This is a framework for building knowledge graphs over complex professional or academic literature: medical papers, legal documents, technical specifications. Every assertion in the graph traces back to a source document and location. Provenance is not an afterthought -- it is load-bearing.

For the conceptual foundation, see Manifesto, Overview and Architecture. For book-length treatments of the full pipeline, see Graphwright Publications.

Live tools

Concepts and Architecture

  • Overview -- why knowledge graphs, core concepts, two-pass ingestion model
  • Architecture -- component breakdown, module structure, immutability design

Identity and Entity Resolution

  • Canonical IDs -- entity lifecycle, authority lookup, synonym caching
  • Deduplication -- merging and flagging near-duplicate entities

Trust and Provenance

  • Provenance -- source attribution, confidence scores, audit trail
  • Conflicting Claims -- representing disagreements, not silently resolving them

Ingestion Pipeline

  • Pipeline -- stages and interfaces (parser → extractor → resolver → embedder)
  • Chunking -- document segmentation strategies
  • Error Handling -- partial extraction, retries, fallback behavior

LLM Integration

Schema Design

Storage, Export, and Serving

Examples