Semantic Extraction and Annotation of Organizational Data: A Multi-Agent Approach for University Environments

A semantic multi-agent system for organizational memory management suited to a university setting is presented in this article. To extract, annotate, and query heterogeneous data, the system uses Semantic Web technologies (SPARQL-Generate, OWL, and SPARQL) and FIPA standards. It is organized into three interconnected modules: user-oriented coordination and interconnection, and semantic knowledge management. It is backed by a hybrid storage infrastructure (Jena Fuseki/GraphDB RDF triplestore and MongoDB NoSQL database). For user interaction and semantic extraction, the architecture uses JADE to coordinate agents. The process uses a semi-automated SPARQL-Generate pipeline to create RDF graphs from HTML, JSON, and XML data. Comparison with XPath/XSLT, XWRAP, W4F, and machine learning techniques highlights the superior semantic reasoning and interoperability. With 93.1% precision, 94.2% recall, 93.6% F1-score, 1.4 s/record processing, and 50% RDF size reduction in comparison to Apache Jena and Protégé, as well as 96% ontology completeness and 91.5% inference relevance, experimental validation on university student record management shows outstanding performance. Robust agent coordination and an easy-to-use interface are features of the implementation. The system emphasizes flexibility, scalability, and interoperability while pointing out its initial complexity and reliance on structured data; future improvements include machine learning integration and expanded use cases like scholarship administration.