A NOVEL CONCEPTUAL FRAMEWORK FOR ENHANCING ACADEMIC DOCUMENT CLASSIFICATION VIA DOMAIN-SPECIFIC FEATURE ENGINEERING IN INSTITUTIONAL KNOWLEDGE SYSTEMS

Abstract

Academic document classification remains a critical challenge for institutional knowledge systems due to the structural complexity and disciplinary diversity of scholarly texts. This paper proposes NFEAD (Novel Framework for Enhancing Academic Document Classification), a conceptual framework designed to guide the development of adaptable, domain-sensitive classification systems. NFEAD is structured around four core pillars: data curation and preprocessing, domain-driven feature engineering, a modular ensemble-based architecture, and evaluation and deployment strategy.The framework emphasizes the use of format-specific parsing tools, noise reduction techniques, and metadata enrichment to prepare heterogeneous academic corpora. It proposes a hybrid feature extraction approach that integrates structural (e.g., section hierarchies, citation graphs), contextual (e.g., metadata embeddings), and semantic features (e.g., ontology-driven topic modeling). These features are mapped into a two-stage modular architecture comprising specialized submodels and an integrative meta-learner, designed for institutional flexibility and scalability. While empirical testing is reserved for future work, NFEAD outlines a roadmap for evaluation, reproducibility, and real-world deployment. This conceptual contribution lays the foundation for future implementation, offering a strategic model for improving classification workflows and enhancing the discoverability of academic content in digital knowledge environments.