PROV-O Ontology

PROV-O defines a foundational data model built on three primary starting point classes and their interrelationships:

• Entity: A physical, digital, or conceptual thing about which provenance is recorded.
• Activity: An action or process that occurs over a period of time and acts upon or with entities.
• Agent: A person, piece of software, or organization that bears some form of responsibility for an activity or entity. These classes are linked by properties such as wasGeneratedBy, used, wasAttributedTo, and wasDerivedFrom to form a directed acyclic graph of lineage.

As an OWL2 ontology, PROV-O enables automated reasoning and logical inference beyond simple data serialization. This formal semantics allows machines to:

• Infer implicit provenance relationships that were not explicitly stated.
• Validate the logical consistency of a provenance graph.
• Integrate provenance data with other domain-specific ontologies using standard semantic web linking. The ontology defines strict domain and range constraints on properties, ensuring that provenance assertions are computationally verifiable.

PROV-O is designed as a generic upper-level ontology intended for extension. Domain-specific applications can create subclasses and sub-properties to capture specialized provenance without breaking interoperability.

• A scientific workflow system might define ex:SimulationRun as a subclass of prov:Activity.
• A content licensing platform could define ex:copyrightHolder as a sub-property of prov:wasAttributedTo. This mechanism ensures that core PROV queries remain valid across all compliant systems while allowing for rich, domain-specific detail.

The standard specifies a mapping to RDF/XML, Turtle, and JSON-LD serializations, ensuring that provenance graphs can be exchanged between disparate systems without loss of fidelity. This is critical for:

• Cross-organization data sharing: Allowing a data publisher and consumer to exchange a verifiable chain of custody.
• Toolchain integration: Enabling a data catalog, an ETL scheduler, and a compliance auditor to all consume the same provenance stream. The JSON-LD serialization, in particular, facilitates integration with modern web APIs and document databases.

PROV-O introduces the concept of a Bundle, a named set of provenance statements that can itself be the subject of provenance assertions. This enables:

• Provenance of Provenance: Recording who asserted a specific lineage record and when.
• Contextual Attribution: A single entity can have different, potentially conflicting, provenance records in different bundles, each attributed to a different source.
• Modularity: Large provenance graphs can be partitioned into manageable, self-contained bundles for distributed processing and storage.

Beyond simple binary relations, PROV-O provides qualified patterns that reify a relationship into a class. For example, instead of a direct wasAttributedTo link, a prov:Attribution instance can be created with its own properties, such as:

• prov:agent: The responsible agent.
• prov:entity: The attributed entity.
• prov:hadRole: A specific role the agent played (e.g., 'editor', 'creator', 'reviewer'). This pattern allows for rich, n-ary relationships that capture the nuanced context of responsibility without breaking the RDF data model.