How cairnepal/hg-skill works
HeritageGraph Skills turns heritage documents, text, and tabular records into ontology-grounded RDF Turtle. Users ingest and preprocess data, extract and validate a LinkML Container, resolve entity URIs (optionally with curator review), map resolved records to RDF, then run SHACL validation. The README documents this staged workflow and run artifacts; sampled code shows selected ingestion, scanning, validation, resolution, mapping, and workspace behavior. The pipeline runner and SHACL validator are listed but unsampled, so their internal calls are not inferred.
Main entry points: run_workspace.py Run Workspace · HeritageGraph.shacl.ttl Quality Shapes
Components 26
Ingestion and Extraction
- Raw Ingestor
- PDF Ingestion
- Tabular Ingestion
- Heritage Scanner
- Container Extraction
- Extraction Validation
- Extracted Container
Entity Resolution
- Resolution Runner
- Gazetteer Matching
- Getty Lookup
- LUX Lookup
- Wikidata Lookup
- Review Queue
- Resolved Records
RDF Mapping
- RDF Mapper
- Mapping Rules
- RDF Graph
Validation and Output
- SHACL Validation
- Quality Shapes
schemas/HeritageGraph.shacl.ttl - Validation Report
- Validated RDF
Run Management
- Pipeline Runner
- Run Workspace
skills/common/run_workspace.py
Other components
- Heritage Sources
- External Authorities
- Human Curator
Connections 35
- Heritage Sources → Raw Ingestor: ingests
README.md - Raw Ingestor → PDF Ingestion: dispatches
- Raw Ingestor → Tabular Ingestion: dispatches
- Raw Ingestor → Heritage Scanner: scans with
- Raw Ingestor → Extracted Container: prepares input
README.md - Raw Ingestor → Run Workspace: records artifacts
- Extracted Container → Container Extraction: feeds
README.md - Container Extraction → Extraction Validation: validates with
README.md - Extraction Validation → Extracted Container: validates
- Extracted Container → Resolution Runner: resolves
README.md - Resolution Runner → Gazetteer Matching: matches with
- Resolution Runner → Getty Lookup: queries
- Resolution Runner → LUX Lookup: queries
- Resolution Runner → Wikidata Lookup: queries
- Getty Lookup → External Authorities: queries
- LUX Lookup → External Authorities: queries
- Resolution Runner → Review Queue: stages candidates
- Human Curator → Review Queue: reviews
README.md - Review Queue → Resolved Records: curates matches
- Resolution Runner → Resolved Records: writes
README.md - Resolved Records → RDF Mapper: maps
README.md - RDF Mapper → Mapping Rules: uses rules
- RDF Mapper → RDF Graph: emits
README.md - RDF Graph → SHACL Validation: validates
README.md - SHACL Validation → Quality Shapes: checks against
README.md - SHACL Validation → Validation Report: writes report
README.md - SHACL Validation → Validated RDF: conforms as
README.md - Pipeline Runner → Resolution Runner: runs stages
README.md - Pipeline Runner → RDF Mapper: runs stages
README.md - Pipeline Runner → SHACL Validation: runs stages
README.md - Pipeline Runner → Run Workspace: uses runs
- Run Workspace → Extracted Container: tracks artifacts
README.md - Run Workspace → Resolved Records: tracks artifacts
README.md - Run Workspace → RDF Graph: tracks artifacts
README.md - Run Workspace → Validation Report: tracks artifacts
README.md