What is SHACL?

A knowledge graph is only as useful as the data in it. Without checks, a missing property, a value of the wrong type or a link to the wrong kind of thing goes in unnoticed, and turns up later as a wrong number in a report or a confused answer from an AI system.

SHACL turns quality rules into something a machine can run. New data is validated as it arrives, and the result is a report of exactly which records broke which rule. Data quality no longer depends on someone reviewing records by hand.

The report is also a record you can show an auditor: which rules ran, against which data, and what failed. That matters in regulated work, for example for risk data under BCBS 239 in banking, the accuracy principle in the GDPR, or data integrity in clinical trials.

A SHACL shape picks out a set of nodes, for example every instance of ex:Person, and lists the constraints those nodes must meet:

Cardinality. Every Person has exactly one name (sh:minCount 1; sh:maxCount 1).

Datatype. A birth date must be an xsd:date (sh:datatype xsd:date).

Pattern. An email address must match a regular expression (sh:pattern).

Structure. Shapes can refer to other shapes, so you can require that every Order links to at least one LineItem and that each LineItem has a positive quantity.

Running validation produces a validation report. Each result names the node that failed, the property and constraint involved, and a message written for people, so whoever owns the data can see what to fix without reading the shapes.

Finance. A bank with a FIBO-based knowledge graph can validate every incoming data feed against its shapes, catching a missing LEI or an entity filed under the wrong classification before it reaches the risk models.

Manufacturing. A manufacturer with plants in several countries can use one set of shapes as its product data standard. When a plant submits a product record without a required attribute, the report says which record and which attribute.

Open data. DCAT-AP, the EU profile for describing datasets, publishes SHACL shapes with the specification. Publishers and portals check dataset descriptions against them, so metadata from different member states follows the same rules and can be searched together.