Skip to content

PROV-O

di# FAIR² and PROV-O Integration

Overview

FAIR² (FAIR Squared) adopts the PROV-O (Provenance Ontology) standard to support structured provenance metadata. This enables datasets to be:

  • Findable: by tracking their origins and transformation processes.
  • Accessible: by providing machine-readable provenance records.
  • Interoperable: through the use of linked data for provenance documentation.
  • Reusable: by enabling transparency and reproducibility in AI and machine learning workflows.

PROV-O is a W3C recommendation for representing provenance information, including data lineage, authorship, and transformations. FAIR² incorporates PROV-O in JSON-LD metadata to improve dataset transparency and facilitate auditability in responsible AI.

PROV-O Concepts in FAIR²

The following table summarizes how FAIR² utilizes core PROV-O terms:

PROV-O Term Purpose in FAIR² Example
prov:Entity Represents a dataset or data file "@type": "prov:Entity"
prov:Agent Identifies individuals, organizations, or software "@type": "prov:Agent"
prov:Activity Describes processes such as data generation or curation "@type": "prov:Activity"
prov:wasGeneratedBy Links a dataset to the activity that created it "wasGeneratedBy": { "@type": "prov:Activity", "name": "Collection" }
prov:wasAttributedTo Associates a dataset with its creator or maintainer "wasAttributedTo": { "@type": "prov:Agent", "name": "Research Lab" }
prov:wasDerivedFrom References prior datasets used in derivation "wasDerivedFrom": "https://doi.org/10.1234/original-dataset"

Example: JSON-LD with PROV-O Metadata

{
  "@context": [
    "https://www.w3.org/ns/prov",
    "https://fair2.ai/ns/"
  ],
  "@type": "Dataset",
  "name": "AI-ready Dataset",
  "description": "A dataset aligned with PROV-O for provenance tracking.",
  "author": {
    "@type": "prov:Agent",
    "name": "Dr. Jane Doe",
    "affiliation": {
      "@type": "Organization",
      "name": "AI Research Lab"
    }
  },
  "wasGeneratedBy": {
    "@type": "prov:Activity",
    "name": "Dataset Preprocessing",
    "startTime": "2025-01-15T10:00:00Z",
    "endTime": "2025-01-15T12:00:00Z"
  },
  "wasAttributedTo": {
    "@type": "prov:Agent",
    "name": "AI Research Lab",
    "role": "Data Curator"
  },
  "wasDerivedFrom": "https://doi.org/10.1234/original-dataset"
}

Rationale for Using PROV-O

Integrating PROV-O into FAIR² enables:

  • Provenance traceability: recording who created, modified, and curated datasets.
  • Reproducibility: documenting data transformations and source dependencies.
  • Compliance with Open Science and Responsible AI practices.
  • Interoperability: aligning with existing W3C standards for metadata reuse.

Next Steps

To include provenance metadata using PROV-O in FAIR²:

  1. Refer to the FAIR² Schema for relevant properties.
  2. Incorporate PROV-O terms into your dataset JSON-LD metadata.
  3. Validate using SHACL rules.
  4. Contribute examples or feedback via GitHub Issues or feedback@fair2.ai.