The exponential growth of digital libraries poses a challenge to information retrieval, often resulting in information overload and siloed reading. This work proposes a multidisciplinary framework that bridges Library Science and Computer Engineering to enhance scientific paper recommendation. By integrating controlled vocabularybased semantic analysis with graph-based centrality metrics, the proposed solution addresses traditional limitations such as the coldstart problem and lack of diversity. The framework utilizes a hybrid filtering approach, combining content-based and collaborative methods with advanced bibliographic heuristics. This integration aims to facilitate interdisciplinary knowledge transfer and improve the precision of literature reviews in STEM fields. This framework addresses the limitations of traditional approaches by incorporating semantic analysis and graph-based metrics, showing superior accuracy (89.05%) with hybrid metadata and robust semantic coverage for cold-start scenarios. The inclusion of ontologies and contextual metadata significantly improves collaborative filtering, reinforcing the importance of semantics for performance.