type: "Research"
title: "Open data for the Explore OKF pilot"
description: "Authoritative UK sources and a bounded Coventry citizen journey for testing readable labels, useful semantic linking and exploratory publication."
tags: [okf, authoring, linked-data, uk-government, public-services]
language: en-GB
generated: { by: process:okf-authoring-methodology-review, at: 2026-08-12T00:00:00Z }
status: draft
sources:
  - { id: w3c-dwbp, resource: "https://www.w3.org/TR/dwbp/" }
  - { id: cpsv-ap, resource: "https://semiceu.github.io/CPSV-AP/" }
  - { id: govuk-content-api, resource: "https://content-api.publishing.service.gov.uk/" }
  - { id: ons-digital-boundaries, resource: "https://www.ons.gov.uk/methodology/geography/geographicalproducts/digitalboundaries" }

Open Data For The Explore OKF Pilot

Status: researched shortlist for owner review, 12 August 2026.

This note selects public, understandable data that can test the revised OKF authoring methodology recorded in the companion documentation source docs/okf-authoring-methodology-review-2026-08-12.md before the full okf-uk-living pack is reviewed. It is a source plan, not a claim that every source has already been acquired or licensed for every use.

Recommended Pilot

Use one bounded citizen story: moving to Coventry and finding everyday local services. Limit the first snapshot to one local-authority area, a small set of service types and roughly 30–60 human-facing entities.

The pilot should let someone answer:

• What official guidance applies when I move home?
• Which council area and boundary contains the selected place?
• Who manages household waste and what official statistics describe it?
• Which schools, health organisations, transport stops and police services are nearby or serve the area?
• Which organisation provides each service, what is its official identifier and where can I verify it?
• Which links are official facts, reviewed mappings, editorial navigation or unresolved hypotheses?

This slice is deliberately smaller than the life-course corpus. It exercises the hard cross-linking problems without repeating Land Registry or processing thousands of records before a person can review the model.

Source Shortlist

Being listed in the UK Government API Catalogue does not prove that an endpoint is open, anonymous, current or licensed for the intended operation. Each source retains its own access and rights decision.

Semantic Model To Exercise

Entity Families

• citizen life event or task;
• public service;
• guidance/content item;
• public organisation and organisational unit;
• dataset and distribution;
• place, statistical geography and boundary representation;
• school, health organisation, transport stop and police neighbourhood;
• concept scheme, concept and mapping assertion; and
• source observation, generation activity, rights statement and evidence item.

Reused Vocabularies

• CPSV-AP for evidenced public services, events, channels and competent authorities;
• DCAT 3 and DCAT-AP where applicable for datasets, catalogues and distributions;
• SKOS for the life-event/topic scheme and qualified mappings;
• PROV-O and Dublin Core Terms for source and generation evidence;
• Schema.org only where an exact class/property mapping is evidenced; and
• GeoSPARQL or an explicitly scoped geospatial projection for geometry and spatial relations, without inferring administrative containment from a rough display polygon.

The W3C Data on the Web Best Practices provides the cross-domain publication baseline: stable identifiers, reused vocabulary, coverage, provenance, licensing, feedback and complementary human/machine presentations.

Cross-Link Register

The pilot should create and review link sets rather than adding incidental edges.

Each row expands into an exact unique candidate-ID inventory plus linked, unresolved, excluded and conflicting candidate-ID sets whose disjoint union is the inventory. The deterministic extraction rule, canonical inventory digest and evidence bind it to the frozen source snapshot. Link assertions are counted separately; each has a stable ID and one identity-bound dereference result. An exclusion names the exact eligible candidate IDs, rule and evidence and cannot overlap another outcome. The pilot should not set a percentage target before the eligible population is known, but graph-reachable label coverage is always 100 per cent. This proves reconciliation only inside the author-declared inventory; owner/domain review must judge whether the deterministic eligibility rule omitted anything from the frozen source snapshot. Mapping predicates must agree with their declared strength. Every target must remain inside its URI-aware namespace, with encoded path delimiters rejected. Duplicate candidate-target assertions are rejected, and dereference outcomes come from machine-readable terminal results. For approved v1 results, one approval-grade evidence item must carry both the canonical complete-result digest and the result's exact observed_at; coverage must also be current.

Boundary Sidecar Test

The Coventry example is well suited to the proposed governed multi-resolution geospatial pack. Cache the same official boundary at the resolutions needed for national overview, regional context and local inspection. Each member records:

• official geography code and boundary vintage;
• source product and retrieval evidence;
• simplification/generalisation level and intended zoom range;
• coordinate reference system and generated GeoJSON projection;
• licence and attribution;
• geometry hash, byte count and bounding box; and
• a rule prohibiting the geometry from being used as more precise evidence than its source/generalisation supports.

This directly tests the earlier failure where a crude three-polygon UK outline looked authoritative at the wrong scale.

Exploratory Publication Shape

The first Explore OKF snapshot should contain:

• a one-page citizen narrative and five to ten competency questions;
• 30–60 entity records with complete readable labels;
• a compact label/type index covering every relationship endpoint;
• a small SKOS concept scheme for “moving home and local services”;
• CPSV-AP service projections only for evidenced services;
• several source-native official identifier mappings;
• a multi-resolution Coventry boundary sidecar;
• link-coverage and unresolved-mapping tables;
• a limitations and source-rights page;
• machine-equivalent YAML-LD/JSON-LD/RDF projections where selected; and
• an exploratory descriptor/banner with route-preserving feedback.

It should not contain the complete okf-uk-living corpus, personal addresses, an inferred legal-advice graph, live credentials or bulk crime-event data.

Test Matrix

Selection Decision

Start with GOV.UK content, ONS geography, WasteDataFlow and a very small NaPTAN/GIAS/NHS ODS selection. Add Police organisation metadata only if the first slice remains easy to understand. Defer Companies House and legislation until the core label, link, banner and feedback contracts pass.

This source order minimises access friction while providing heterogeneous identifiers, organisations, services, datasets, places, geometry and concept mappings. It is sufficiently everyday to explain an ontology to a novice and sufficiently structured to expose authoring-method defects before the okf-uk-living review.