Data quality assurance and maintenance were addressed through structural/semantic placement checks and similarity-based deduplication. Cross-corpus similarity analyses were performed over (i) Scrambl’s proprietary database, (ii) the database produced in this study, and (iii) the database previously developed for Scrambl by Koller, 2025. Threshold-based vector similarity identified duplicate and near-duplicate skill entries; the ensuing consolidation reduced redundancy and improved consistency across corpora. The resulting dataset constitutes an enterprise-ready catalog of education programs and their machine-generated skill sets, whose alignment with source descriptions was verified.

Overall, the research delivers a reproducible pipeline for KG construction and upkeep - from automated skill extraction and graph ingestion to systematic deduplication - supporting explainable, skill-gap-aware recommendations at scale. The artifact and its evaluation demonstrate that dynamically generated, quality-assured skill sets can be sustained within a graph-based architecture suitable for real-world workforce development contexts.