A Comprehensive Review of Methods, Frameworks, and Domains for Metadata Extraction from Scientific Texts

Metadata extraction from scientific texts plays a crucial role in enabling efficient organization, retrieval, and analysis of scholarly knowledge. With the exponential growth of scientific publications across disciplines, manual metadata annotation has become infeasible, motivating the development of automated and semi-automated extraction techniques. This review paper presents a comprehensive analysis of recent advances in metadata and structured information extraction from scientific documents. It explores traditional rule-based methods, machine learning approaches, deep learning architectures, and emerging large language model-based frameworks. The paper also examines domain-specific applications, including systematic reviews, digital libraries, scientific repositories, and open journal systems. By synthesizing findings from recent literature, this study highlights key research trends, strengths, and limitations of existing methods. Furthermore, it identifies major challenges such as document heterogeneity, semantic ambiguity, evaluation complexity, and human–machine collaboration. The review aims to provide researchers with a structured understanding of current methodologies and open research directions, thereby supporting the development of robust, scalable, and high-precision metadata extraction systems for scientific knowledge management.