dpv
Theme
Text size
GitHub Menu

dpv Knowledge Bundle

Data Privacy Vocabulary (DPV) 2.3 (LinkML)

Knowledge bundle LOKF 0.2https://w3id.org/lmodel/dpv/knowledge/

Health50 concepts

Licence
CC BY 4.0
Publisher
Noel McLoughlin
As data
concepts.jsonld graph.json graph.nt datasets.jsonld

A LOKF knowledge base for dpv. Every Markdown file under knowledge/ is one concept; together they form a queryable knowledge graph, derived from this repository’s code and docs.

Services#

  • The converter - scripts/dpv_to_linkml.py: reads the Turtle under upstream/ and writes the terms module, the kinds’ modules, the catalogue, the examples and the SSSOM sets. just gen-schema runs it, and it stops on a predicate no table names.
  • Documentation site - The knowledge website in website/, a Hugo site that the Knowledge website workflow publishes to GitHub Pages at https://lmodel.github.io/dpv, built from the README, docs/, the schema’s pages and this bundle.

References#

  • Data Privacy Vocabulary (DPV) - The upstream authority: DPV 2.3 of 2026-02-28, the vocabulary of the W3C Data Privacy Vocabularies and Controls Community Group for describing the processing of personal data, at https://w3id.org/dpv/2.3.
  • The DPV repository - The Community Group’s repository, w3c-cg/dpv, whose tag dpv-2.3 names commit cbcd215. It keeps each edition in a folder, and the examples and the mappings at its root.

Datasets#

  • The DPV LinkML schema - src/dpv/schema/: the umbrella dpv.yaml over dpv_core, dpv_terms and one module of kinds per namespace family. It models DPV 2.3 under DPV’s own IRIs, and DpvCatalogue is its tree root.
  • The core module - src/dpv/schema/dpv_core.yaml, written by hand: the date and address types, the term status list, the classes for a vocabulary, a concept scheme, an agent, a web page and an interval, and the slots they share.
  • The terms module - src/dpv/schema/dpv_terms.yaml, which the converter writes: DpvTerm and its children DpvConcept, DpvProperty and DpvResource, DpvExample, DpvStatement, and one slot per DPV property under its own IRI.
  • The kinds’ modules - The modules the converter writes with one class per kind of term, 149 in all, under each kind’s own IRI: dpv_kinds for the core, and dpv_pd, dpv_tech, dpv_ai, dpv_risk, dpv_legal, dpv_sector and dpv_standards.
  • The catalogue - src/dpv/data/: one YAML file per namespace family, with every vocabulary, concept scheme, concept and property of DPV 2.3 as a record under its own IRI: 8,856 concepts and 307 properties.
  • The examples as data - src/dpv/data/dpv-2.3-examples.yaml: DPV’s 88 examples, each with its description, and the statements of 85 of them as DpvResource records. Three files do not parse.
  • Where DPV’s examples leave a property’s usual classes - tests/example_departures.yaml: each place DPV’s own examples use a property beyond its rangeIncludes or domainIncludes, or give a literal where the range names a class: 58 values, 65 subjects and 30 literals.
  • The contributors - src/dpv/data/dpv-2.3-agents.yaml: the 70 people and organizations DPV’s files name as contributors and creators, each with an id of the form dpv_agent:person-<name> or dpv_agent:org-<name>.
  • DPV’s own mappings as SSSOM - src/dpv/mappings/: two SSSOM sets that render the mapping files DPV publishes, 5 rows for ISO/IEC 29100:2024 terms and 24 for ODRL, each row reading with DPV’s term as its subject.
  • The copy of DPV’s files - upstream/: the Turtle of each vocabulary, the 88 examples, the ISO and ODRL mapping files and the licence, unchanged at the tag dpv-2.3. scripts/upstream_pins.tsv holds each of the 160 files’ git object hashes.
  • The adapt copy - .lokf/schema/dpv.yaml, which just lokf-adapt writes from the schema so that it can sit beside LOKF’s: the file this bundle validates against, and the file a release attaches as schema-<tag>-dpv.yaml.
  • The neighbours’ copies - tests/neighbours/: the adapt copies 27 other lmodel repositories release, each the release’s schema asset, which tests use to hold this copy’s names apart from theirs.
  • The example data - tests/data/: a valid example per class, each kind’s a real record of the catalogue, a counter-example per constraint the validator can state, and one per rule it cannot, under problem/invalid/.
  • The generated artefacts - What the generators write from the schema: project/, src/dpv/datamodel/ and docs/schema/dpv.yaml. They are committed and never edited, and the SQL generator is off.
  • Marketing - The purpose dpv:Marketing of DPV 2.3, written as a concept of type DpvPurpose in the adapt copy’s keys. It shows that the domain schema admits a term of a kind.
  • A process with a purpose - The process exd:PDH1 of DPV’s example E0026, written as a concept of type DpvResource. It shows that the domain schema admits a thing a use case describes, with any DPV property.

Explanations#

  • Why terms are records - Why the schema holds DPV’s terms as records and not as classes: DPV states each concept as a class and as an instance, so one class per term could not say that a process’s hasPurpose is dpv:Marketing.
  • How kinds and slots get their names - How the converter names what it writes: a kind’s class takes a Dpv prefix and a slot keeps its property’s local name, and where two namespaces define one name, the core keeps it and the other takes its namespace first.
  • What stays open - Why no validator refuses a value for leaving a property’s usual classes: DPV states domainIncludes and rangeIncludes, not rdfs:domain and rdfs:range, so DpvResource holds any property.
  • What the records leave out - The facts of DPV’s files no record holds, each with its reason: skos:narrower, which the records regenerate, the repeated schema: domain and range, what says how the site publishes a file, and dpv-27560.
  • What the data shows - What DPV’s own files hold that a reader would not expect: names defined nowhere, predicates no vocabulary has, three example files that do not parse, and a term with two creation dates.
  • The mappings - Which mappings this repository holds and why: only what DPV’s editors publish in machine-readable form, for ISO/IEC 29100:2024 terms and for ODRL. The other 24 sets, which a matching tool wrote, are gone.

Glossary#

  • Term - A term: one thing DPV’s vocabularies state as a concept or a property, under its own IRI. Each is one record here, a DpvConcept or a DpvProperty.
  • Kind - A kind: a type a DPV concept holds beyond rdfs:Class and skos:Concept, such as dpv:Purpose. Each of the 149 kinds is a class of this schema, and a record’s rdf_type names its kinds.
  • Resource - A resource: a thing a use case describes with DPV, such as a process. DpvResource holds any DPV property, because DPV states none as limited to a class.
  • Namespace family - A namespace family: the vocabularies this project keeps in one data file, such as legal, which holds the law vocabulary of each jurisdiction and each EU law.
  • Departure - A departure: a place DPV’s own examples use a property beyond what its rangeIncludes or domainIncludes name, or give a literal where the range names a class. It breaks no rule of RDF.
  • Adapt copy - An adapt copy: the form of a schema lokf adapt writes so that it can sit beside LOKF’s, in which concept classes descend from Concept and keys LOKF also has are renamed. Here it is .lokf/schema/dpv.yaml.

Playbooks#

  • Regenerate and test - How to change the schema: edit the converter or a hand-written module, run just gen-schema, just lint, just gen-project, just gen-doc, just lokf-adapt and just test, and commit what they write.
  • Validate a file of the catalogue - How to validate a file: uv run linkml-validate -s src/dpv/schema/dpv.yaml -C DpvCatalogue <file>, or the committed JSON Schema without LinkML, and tests/dpv_rules.py for what a validator cannot state.
  • Describe a use case - How to describe a thing in a use case with DPV: write a DpvResource with an id, the term it is an instance of in rdf_type, and DPV’s properties as keys, then validate it with -C DpvResource.
  • Update the upstream pins - What to do when the watch reports another DPV release: read what changed, replace upstream/, the pins and the headers together, regenerate, and look at each test that fails.
  • Cut a release - How a release happens: write ## [Unreleased], merge with a releasing commit type, and semantic-release.yml tags, writes the changelog and dispatches the knowledge release, which attaches the assets.
  • Knowledge sources map - The librarian’s scrape map: every repository location and external source this bundle derives concepts from, how to re-check each, and the tracked paths left out with their reasons.

Policies#

  • AI covenant - The rules for AI-assisted work here: a person owns every contribution, an element says what DPV says, a rule states only what DPV states, and a mapping is DPV’s own or it is absent.
  • Licence and attribution - Whose work each part is: the code and the schema under Apache-2.0, the bundle under CC BY 4.0, and DPV’s files and the text taken from them under the W3C Software and Document License, which NOTICE reproduces.
  • Security policy - How to report a vulnerability, what this repository releases, what executes here and what holds each surface. A report goes through GitHub’s private vulnerability reporting, and only the latest release receives fixes.

Workflows#

  • Build and test (main.yaml) - The Build and test workflow: on every pull request and push to main, it runs just test on Python 3.11 to 3.14, and lints the schema and checks the adapt copy on 3.13.
  • Lint and docs (lint-and-docs.yaml) - The static checks on pull requests and on main: ShellCheck, actionlint, the governance files’ word budgets, markdownlint, lychee and codespell.
  • Semantic release (semantic-release.yml) - The workflow that computes the next version from Conventional Commits on main, promotes the changelog, tags and publishes the GitHub Release, and dispatches the knowledge release.
  • Upstream watch (upstream-watch.yaml) - The weekly check that scripts/check_upstream_release.py runs: another release of DPV, a moved tag, a changed or missing pinned file, or a changed licence opens or updates an issue.
  • Knowledge registrar (knowledge-registrar.yaml) - The gate on the bundle: it validates the bundle on each pull request that touches it, and checks who vouched for what.
  • Knowledge librarian (knowledge-librarian.yaml) - The sidecar’s scheduled agent run that keeps the bundle in step with the repository. It stays inert until KNOWLEDGE_LIBRARIAN_ENABLED is true, and a person merges what it proposes.
  • Knowledge release (knowledge-release.yaml) - The workflow that attaches the bundle, a Copilot skill and the schema’s adapt copy to a GitHub release, each with its .sha256 and an attestation, when the bundle or the schema changed.
  • Knowledge sync (knowledge-sync.yaml) - The weekly workflow that keeps the sidecar’s copies and the skills pin at the newest release of Knowledge Trust Ladder, by a pull request a person merges.
  • Knowledge website (knowledge-website.yaml) - The workflow that builds website/ on each pull request that touches its inputs, and deploys it from main once KNOWLEDGE_WEBSITE_ENABLED is true.