Formalizing Feedback as Reusable Knowledge in Data Preparation Systems: A Persistent Feedback Artefact Model

Feedback-driven data preparation systems use user corrections to improve data quality and transformation accuracy. However, existing approaches typically treat feedback as a transient interaction that is consumed during execution and discarded after task completion. This limits the reuse of corrective knowledge across datasets, users, and preparation cycles. This study proposes a Persistent Feedback Artefact Model (PeFAM) as a conceptual and formal representation for treating feedback as a reusable computational knowledge object. The model represents feedback as a structured artefact comprising the original value, corrected value, contextual attributes, metadata, timestamp information, and provenance. A context-aware similarity mechanism is defined to guide how feedback artefacts may be retrieved and assessed for reuse across preparation tasks. The model was assessed through a worked analytical scenario that examined representational completeness, traceability, contextual matching, and threshold-based eligibility for human review. The study contributes a formal representation of feedback knowledge, a similarity-based reuse mechanism, and a conceptual arrangement of functions supporting persistent feedback management in data preparation environments. The analytical demonstration indicates that the model can support representation, traceability, contextual comparison, and threshold-based eligibility within the defined scenario, while empirical evaluation of efficiency, accuracy, usability, and scalability remains necessary.