Research results document
A research results document is not a thesis or an essay; it is a structured, auditable record of findings that a later consumer must be able to trust and reuse. Capture provenance, not prose.
Per-claim record
Each claim MUST carry:
- The claim stated as one verifiable sentence.
- At least one citation with: title, author or organization, publisher or venue, date, DOI or URL, and the access date.
- The resource / source type (peer-reviewed, official docs, government data, clinical trial, news, reference, …).
- The exact supporting quote — verbatim, never paraphrased.
- Who or what generated the claim, and the query used to find it.
- A trust / confidence score and the basis for it.
- The independence of its support (corroboration: multi-independent, single, uncorroborated).
Document container
The document SHOULD record: topic and scope; the research questions; methodology (search strategy and inclusion/exclusion); the agents and sources involved; dates; an executive summary; the findings; open questions; limitations; and license.
Provenance standards to align with
- W3C PROV — entity / activity / agent, with
wasDerivedFrom,wasAttributedTo,hadPrimarySource. - PAV —
retrievedFrom,retrievedOn,createdWith,version. - FAIR — persistent identifiers (F1) and detailed provenance (R1.2).
- schema.org Claim / ClaimReview — for fact-check-style records.
- DataCite / Dublin Core / Crossref — for citation metadata.
- Nanopublications — the assertion / provenance / publication-info pattern for atomic, attributable claims.
- Datasheets for Datasets / Data Cards — when the result is itself a dataset.