Intelligent Metadata Architecture for Digital Institutional Repositories in Bangladesh: A Semantic Web Approach to Knowledge Organization and Discovery

Abstract

Background: Institutional repositories (IRs) have emerged as critical components of academic knowledge management infrastructure, enabling universities to collect, preserve, and disseminate scholarly output. In Bangladesh, where higher education institutions are rapidly expanding research capacities, the quality of IR metadata architecture directly determines whether accumulated scholarly knowledge is effectively discoverable, interoperable, and usable by researchers and policymakers. Problem Statement: Existing institutional repositories in Bangladeshi universities predominantly implement flat, keyword-based metadata structures based on Dublin Core standards that lack semantic richness, limit cross-repository interoperability, and fail to capture the relational context essential for advanced knowledge discovery. Proposed System: This paper proposes SMIR-BD, a Semantic Metadata-enriched Institutional Repository architecture for Bangladesh, integrating Linked Data principles, a Bangladesh Academic Knowledge Ontology (BAKO), and a SPARQL-based semantic query engine. Methodology: The system was developed and evaluated using 680 research documents harvested from five Bangladeshi university repositories. Key Findings: SMIR-BD achieved an information retrieval precision of 87.3%, recall of 85.6%, and F1 of 86.4%, compared to 59.4%, 62.6%, and 69.8% respectively for conventional Dublin Core-based retrieval. Applications: The system supports national research knowledge management, academic search platforms, and policy-relevant research synthesis.