W3C PROV

Defines the fundamental building blocks for representing provenance as a directed graph. The model centers on three core types:

• Entity: A physical, digital, or conceptual thing with a fixed state (e.g., a specific version of a document).
• Activity: An action or process that occurs over time and generates or modifies entities (e.g., an editing session).
• Agent: An actor bearing responsibility for an activity or entity (e.g., a person, software service, or organization). These are linked by relations like wasGeneratedBy, used, wasAttributedTo, and wasDerivedFrom to form a complete lineage graph.

Provides a formal semantic web mapping of the PROV data model using the Web Ontology Language (OWL2). This enables:

• RDF Serialization: Provenance data can be expressed as RDF triples and queried using SPARQL.
• Semantic Interoperability: Allows provenance graphs to be linked with other domain ontologies and linked data on the web.
• Logical Reasoning: The formal axioms enable automated inference over provenance statements, such as deducing transitive responsibility chains.

A human-readable, functional-style syntax designed for writing and illustrating PROV instances directly in documents and examples. It serves as the canonical representation in the specification. For example: entity(ex:report-v1) wasGeneratedBy(ex:report-v1, ex:write-activity, 2024-05-20T10:00:00Z) This notation is unambiguous and maps directly to the PROV-DM abstract model, making it ideal for teaching and precise specification.

Defines a set of formal constraints and inference rules that a valid PROV instance must satisfy. These rules ensure data integrity and logical consistency. Key constraint types include:

• Uniqueness Constraints: Ensure that an entity cannot be generated by two different activities.
• Ordering Constraints: Define the temporal logic, such as an entity's generation must precede its usage.
• Typing Constraints: Enforce that relations connect the correct types of elements (e.g., an Agent cannot be used by an Activity). Validation against these constraints is critical for automated provenance processing.

Standardizes the mechanism for discovering and retrieving provenance information for a resource on the web. It specifies:

• HTTP Link Headers: A resource can advertise the location of its provenance record using a rel="http://www.w3.org/ns/prov#has_provenance" link.
• RESTful APIs: Defines a protocol for querying a provenance service by submitting a resource URI to retrieve its provenance graph. This protocol decouples the artifact from its lineage, allowing lightweight resources to link to rich, separately managed provenance stores.

An extension to the core model that introduces the concept of a Dictionary, a special type of Entity with a key-value structure. This is crucial for tracking provenance in data pipelines and scientific workflows where:

• A single activity inserts, updates, or deletes specific members of a collection.
• Fine-grained attribution is needed for individual entries within a large dataset.
• The exact state of a parameter set used to execute a computational activity must be recorded immutably.

The OWL2 Web Ontology Language encoding of the PROV Data Model. PROV-O maps the three core PROV classes—Entity, Activity, and Agent—and their interrelationships into a semantic web vocabulary. This enables provenance assertions to be published as RDF triples, making them queryable via SPARQL and linkable across the open web. It provides a lightweight, domain-agnostic starting point for building interoperable provenance graphs.

A human-readable, functional-style syntax for writing PROV instances. PROV-N is designed to be easily writable by humans while remaining precisely parsable by machines. It expresses provenance as a series of relation expressions, such as wasGeneratedBy(ex:report, ex:analysis, 2024-01-15T10:00:00Z). This notation serves as the canonical textual representation in the PROV specification suite, bridging the gap between formal data models and developer ergonomics.

A formal specification of validity rules and inferences that apply to PROV instances. PROV-Constraints defines a set of logical constraints—including typing, ordering, and impossibility constraints—that a valid provenance record must satisfy. These rules enable automated validation of provenance graphs to detect logical inconsistencies, such as an entity being used before it was generated. It provides the formal underpinning for reasoning engines to infer implicit provenance relationships.

A protocol specification for discovering and retrieving provenance information over HTTP. PROV-AQ defines how a resource can advertise the location of its provenance records using HTTP Link headers with the rel="provenance" relation type. It also specifies how to query a provenance service using a target-uri parameter. This mechanism decouples the artifact from its provenance, allowing lightweight resources to link to rich, separately managed provenance stores.

An extension to PROV that introduces the concept of a Dictionary—a special type of Entity whose structure is a set of key-entity pairs. This enables modeling of complex data structures where individual members can have their own provenance. Key operations include:

• Insertion: Adding a new key-entity pair
• Removal: Deleting a key-entity pair
• MemberOf: Asserting that a specific entity is a member of a dictionary at a given key This is critical for tracking provenance in JSON documents, database rows, or parameter sets.

A PROV construct that allows a set of provenance statements to be treated as a named, first-class Entity itself. A Bundle is a provenance record about provenance—it captures the context in which a set of assertions was made, including the asserting agent and the time of assertion. This mechanism is essential for modeling multi-source provenance aggregation, where different authorities contribute partial views of a complete provenance chain, and the origin of each view must itself be tracked.