Tatiana Orel

This research to address the barriers to integrating Large Language Models (LLMs) into Information Management (IM) within public-sector organizations, primarily stemming from weak Information Governance (IG) and semantic misalignment between structured and unstructured data. The core objective is to map and synthesize diverse semantic concepts across knowledge description and representation systems grounded in information science traditions, data architecture, linguistics, and Artificial Intelligence (AI) governance to develop a unified semantic harmonization layer. This layer would serve as a foundation for responsible and explainable AI by grounding LLMs in the contextual meaning, provenance, and evidential value of information, thereby enabling reliable automation of metadata tagging and classification. Anchored in a post-positivist worldview, this research employs a mixed-methods design that prioritizes qualitative inquiry through conceptual modeling to develop a cross-organizational Human Resources ontology, which is subsequently validated via quantitative performance testing in two public administrations.