A Model-Driven Pipeline for Data Quality Specification and Operationalization: A No-Code Approach for Domain Experts

Data quality is critical for reliable research and decision-making, particularly in fields like cultural heritage where data is often manually curated. While domain experts understand the semantic requirements for high-quality data, they typically lack the technical expertise in database query languages (e.g., SQL, SPARQL, Cypher) required to implement automated quality checks. Conversely, data engineers possess the technical skills but often lack the domain-specific context. This dependency creates bottlenecks, slows down quality assurance, and hinders the ability to adapt to changing data requirements.

The authors propose a four-step, model-driven pipeline designed to decouple quality analysis from specific database technologies. The pipeline is structured into four sub-processes:

1. Requirement Elicitation: Domain experts identify quality issues and define requirements in natural language.
2. Constraint Specification: Experts use the web application 'Constrainify' to select and parameterize existing templates from a library, turning them into executable constraints.
3. Template Creation: Data engineers define new, reusable templates using the Quality Pattern Model (QPM) metamodel when no existing template fits the requirement.
4. Quality Analysis: Data analysts execute the resulting constraints against datasets to generate reports for quality improvement.

The core of this approach is the Quality Pattern Model (QPM), a metamodel that abstracts the logic of quality checks away from specific database implementations. By using QPM, the system can automatically translate a single template into executable queries for different backends, including XML, RDF, and Neo4j. The web application Constrainify provides a user-friendly interface that guides domain experts through the parameterization process, effectively lowering the barrier to entry for performing complex data validation.