Applying The Domain Profile Across Three Collections

These are design checks, not substitutes for a current warm-up. Each new run must inspect the real collection and pin official standards as of its research cut-off.

Comparison

Domain decision UK legislation ONS data discovery GOV.UK content
Typical record boundary Legal work, expression/version, manifestation and provision must remain distinct Dataset/product, release, series, geography and classification may have different identities Content item, edition, route, organisation, taxonomy term and attachment may have different identities
Source authority legislation.gov.uk/TNA for supplied legislation metadata and effects; other legal source owners remain separate ONS source products/catalogues for supplied metadata; reviewed live references remain distinct from governed snapshots Publishing applications and GOV.UK APIs for their supplied fields; navigation observations are not editorial authority
Critical time distinction enactment, commencement/effect, validity, amendment, publication and observation data reference period/release versus catalogue creation/modification and acquisition first publication, public update, edition, route/redirect observation and acquisition
Domain-standard candidates ELI, CLML/LegalDocML, ECLI where source-assigned, Schema.org Legislation, PROV-O SDMX, DDI, statistical classifications, DCAT/CSVW, SKOS, PROV-O GOV.UK content models, Schema.org, DCAT for catalogue-like subsets, SKOS for governed taxonomies, PROV-O
False-equivalence risk Same title or citation does not prove the same work/version/provision Similar product titles do not prove one series or substitute dataset A redirected URL, mirrored page or shared taxon does not prove identical content
Relationship evidence Official amendment/effect data stays separate from model-discovered topics/entities Series membership needs an explicit ID or source-declared scoped label; similarity is presentation-only Browse links, taxon membership, organisation ownership and related-content links retain their distinct source predicates
Likely architecture Large legislation child plus Whole-Law federation for independently governed source families Large static discovery corpus with compact facets and provider datapacks Large corpus; federation only where independently governed products justify it
Model assistance Candidate topics/concepts/entities only, with evidence and independent review Candidate semantic alignment or confusable-alternative discovery, never statistical fact authority Candidate topic/entity/navigation enrichment, never editorial status or policy authority

UK Legislation Assertions

A passing domain profile should:

It should fail review if it creates amendment relationships from title similarity, represents a model topic as official legal classification, or claims complete Whole-Law coverage from legislation.gov.uk alone.

ONS Assertions

A passing domain profile should:

It should fail review if a catalogue metadata_modified value becomes the dataset's latest data period, similar titles alone create a series identity, or metadata completeness is reported as statistical accuracy.

GOV.UK Content Assertions

A passing domain profile should:

It should fail review if all links become one generic relatedTo predicate, a live route is treated as a permanent edition identifier, or a broad GOV.UK coverage claim lacks a declared union of sources and omissions.

Shared Result

The three profiles differ in entities, standards and time semantics but feed the same build controller:

  1. immutable source receipts;
  2. source-native normalization;
  3. OKF 0.2 Markdown;
  4. optional YAML-LD/JSON-LD/RDF with semantic equivalence;
  5. an appropriately sized Explorer data plane;
  6. evidence-bearing relationships and facets;
  7. user/task evaluation;
  8. frozen-candidate assurance; and
  9. identical RC-to-final promotion.