Machine-readable evidence

Data for retrieval and RAG

Use the source-cell JSONL files as retrieval records. Each line is self-contained: a source reporting subject, exact RDF cell identifier, citation URL, workbook checksum, label, reporting geography vintage, variable definition, reference-period status, unit status and original value.

numeric_value retains the staged lexical value; rdf_decimal is the exact decimal string used in the RDF. For blank and text records, rdf_decimal is null. Null years must not be replaced by census vintage.

Chunk by source row or cell. Retrieve the row context along with each cell, preserve qualifications, and cite the citation_url. Do not join a cell to a map using source_code alone. Candidate map matches are deliberately kept outside the cell assertion.

Aggregate source data

Qualified geographic and identity evidence

These files retain assessment methods and statuses. Same-year topology intersections are not a boundary-adjacency network. Cross-year intersections do not assert historical succession.

Wikidata reference links

In wikidata_associations.csv, link_accepted identifies retained reference links. The acceptance basis, verification evidence and evidence gaps explain each assessment. Only specific conflicting evidence creates a review flag, with reasons in review_reasons_json. The review queue contains those flagged links; the QA sample supports periodic checks of accepted links. Link acceptance does not assert identical census boundaries or transfer statistics to a Wikidata entity.

Edition and dataset boundaries

Build manifest and SHA-256 checksums · Interpretation and original sources

The preserved 1881 resident pages and biographical tables are separate retrieval sources. They are not included in the aggregate-cell JSONL distributions.