SHACL

SHACL defines constraints through shapes—blueprints that specify the expected structure and content of RDF nodes. Each shape targets a specific class of resources and declares the properties, cardinalities, and value types those resources must satisfy.

• Node shapes validate individual resources against closed conditions
• Property shapes define constraints on specific predicates
• Shapes can be composed and reused across different validation contexts

A shape for a legal contract might require exactly one ex:effectiveDate, at least two ex:party references, and that every ex:governingLaw value belongs to a controlled jurisdiction vocabulary.

SHACL provides a rich vocabulary of built-in constraint components that express validation conditions without custom code. These components cover the most common data quality checks required in legal knowledge engineering.

• sh:minCount and sh:maxCount enforce cardinality on properties
• sh:datatype restricts literal values to specific XML Schema types
• sh:pattern applies regular expression matching to string values
• sh:class ensures object values are instances of a specified class
• sh:in restricts values to a predefined enumeration

For legal graphs, sh:pattern can validate citation formats like "\d+ U.S. \d+" for US Supreme Court references, catching malformed citations before they enter the knowledge base.

Every SHACL validation produces a machine-readable validation report conforming to the SHACL Validation Report vocabulary. This standardized output enables automated quality pipelines to react programmatically to data issues.

• Each violation is reported as a sh:ValidationResult with severity levels
• Results link directly to the offending node and the violated shape
• Supports sh:Violation, sh:Warning, and sh:Info severity tiers
• Reports can be serialized in Turtle, JSON-LD, or other RDF formats

In a legal document processing pipeline, a Warning on a missing optional clause might trigger manual review, while a Violation on a missing mandatory party identifier could halt ingestion entirely.

SHACL and OWL serve fundamentally different purposes in the semantic technology stack, and understanding this distinction is critical for legal knowledge graph architects.

• OWL operates under the Open World Assumption and focuses on inferring new knowledge from existing assertions
• SHACL operates under the Closed World Assumption and validates that data conforms to expected patterns
• OWL is suited for classification and reasoning; SHACL is suited for data integrity and quality gates
• SHACL can validate data that uses OWL ontologies, but the two standards address orthogonal concerns

For legal applications where missing information is a defect rather than an unknown, SHACL's closed-world validation aligns with the rigor required for contract analysis and regulatory compliance.

Beyond built-in constraint components, SHACL supports SPARQL-based constraints that express arbitrarily complex validation logic. This extensibility allows legal knowledge engineers to encode domain-specific rules that go beyond simple property checks.

• sh:sparql constraints execute custom ASK or SELECT queries
• Can traverse multi-hop relationships in the graph
• Enables cross-referencing between different named graphs
• Supports parameterized constraint templates via sh:SPARQLConstraintComponent

A SPARQL constraint could verify that every ex:Contract referencing a ex:GoverningLaw from a specific jurisdiction also includes a ex:DisputeResolution clause, encoding a complex legal drafting requirement as a single reusable validation rule.

SHACL provides multiple targeting mechanisms that determine which nodes in the graph are subjected to which shapes. This flexibility allows validation scopes to be precisely defined without modifying the data itself.

• sh:targetClass applies shapes to all instances of a specified class
• sh:targetNode targets specific individual resources by IRI
• sh:targetSubjectsOf and sh:targetObjectsOf target nodes based on predicate usage
• Multiple targets can be combined within a single shape

In a multi-jurisdictional legal graph, different shapes can target contracts based on their ex:jurisdiction property, applying Delaware-specific validation rules to Delaware-governed agreements while using different constraints for English law contracts.