Model-Driven Method for Data Quality Assurance

Abstract

Ensuring data quality through validation against structural and semantic constraints defined by a specific use case is critical, mainly when employed to train machine learning models or make accurate analyses and decisions. In this paper, we investigate the usage of Refinery with Graph Queries to define constraints, ensuring consistency between and validating data against them. We will compare the result with other relevant approaches like Ontology with Semantic Web Rule Language (SWRL), Cognipy with Controlled Natural Language (CNL), and Shapes Constraint Language (SHACL). The comparison focuses on structural and syntactic validation, semantic precision, reasoning capabilities, and logic and arithmetic expressions support. The evaluation uses Experience API (xAPI) data, a learning technology interoperability specification designed to track and share learner activity and experiences. Its triple structure (Actor, verb, object) and domain profile features make xAPI data a good fit for validating the proposed approach.