Content provenance

The Citead team, updated

In short. Content provenance records the origin and derivation of information, including the activities and agents responsible for producing it.

Definition: Content provenance records the origin of information and how it is produced or derived from earlier material, including the activities and responsible people, organizations, or software involved. W3C's PROV-O recommendation supplies a vocabulary for representing and exchanging these records between systems.

Key facts

  • W3C publishes PROV-O: The PROV Ontology as a Recommendation dated 30 April 2013.
  • The PROV Model Primer identifies provenance records as metadata and gives web pages and charts as examples of things whose provenance can be described.
  • PROV-DM: The PROV Data Model treats derivation as a relationship between an output and the material from which it originates.
  • The PROV-AQ Working Group Note describes linking a web resource to its provenance record through an HTTP header or an HTML element.

What does a provenance record describe?

The W3C PROV Model Primer calls the things being described entities. A document, a chart within it, and the dataset behind that chart can each be an entity. An activity produces or changes entities, such as translating a document. An agent bears responsibility and can be a person, an organization, or software.

The primer describes complementary views of a publishing process. For a news photograph, provenance can identify the photographer, editor, and newspaper. For an assembled web page, it can trace individual portions to a news article, interview quotations, or government data. A process record can describe retrieving data, calculating statistics, and drawing a chart. The record's detail depends on which of these questions it needs to answer.

How does PROV-O express derivation?

The PROV-O starting point terms connect entities through prov:wasDerivedFrom. They connect an output to its producing activity through prov:wasGeneratedBy, and to a responsible agent through prov:wasAttributedTo.

This shortened extract uses the recommendation's illustrative crime-statistics chart example:

@prefix prov: <http://www.w3.org/ns/prov#> .
@prefix : <http://example.org#> .

:bar_chart
    a prov:Entity;
    prov:wasGeneratedBy :illustrationActivity;
    prov:wasDerivedFrom :aggregatedByRegions;
    prov:wasAttributedTo :derek;
.

The chart derives from aggregated regional data. The identifiers distinguish that input, the illustration activity, and Derek, the responsible agent. The full example records an earlier aggregation activity using government crime data and a separate list of national regions.

How can a website expose provenance?

W3C's PROV-AQ note, also dated 30 April 2013, describes placing a link in a page's HTML head. The following adaptation points to an example provenance document:

<link rel="http://www.w3.org/ns/prov#has_provenance"
      href="https://example.com/provenance/resource.rdf">

The href identifies the provenance record. The note also defines has_anchor for identifying the subject document within that record. Its treatment of changing resources distinguishes a document's current address from identifiers for particular revisions, allowing a history to refer to the version actually described.

FAQ

Can provenance record revisions and quotations?

Yes. W3C's PROV Data Model defines a revision as a derived version containing substantial content from an original. A quotation repeats some or all of an earlier entity, including text or an image.

Can I record who supplies the provenance itself?

Yes. The same data model describes a bundle, a named set of provenance descriptions that can itself have provenance. This allows responsibility for the record to be recorded alongside the history of the content.

Does a provenance record prove that content is accurate?

W3C's PROV-AQ note requires trust in the provenance record to be assessed separately from trust in the original resource. It names the serving domain and an associated digital signature as possible grounds for that assessment.

Sources

  1. W3C, "PROV-O: The PROV Ontology", 30 April 2013
  2. W3C, "PROV Model Primer", 30 April 2013
  3. W3C, "PROV-DM: The PROV Data Model", 30 April 2013
  4. W3C, "PROV-AQ: Provenance Access and Query", 30 April 2013

Check how a page reads to answer engines. Open the GEO score analyzer

Test a page and see what to fix first

The analysis usually takes less than 10 seconds and needs no account.

400,000+ pages scored in the public benchmark