A Review of Automatic Identification and Extraction of Innovation Points in Academic Papers: Progress and Paradigm Shifts

Abstract Purpose This study aims to systematically review the core concepts, technical approaches, research themes, and tool ecosystems related to “academic innovation point extraction”, clarify the developmental trajectory of the field and identify current challenges and future directions. Design/methodology/approach A systematic literature review approach is adopted, supported by BERTopic-based hierarchical topic modeling to identify major research themes. Academic innovation points are categorized into explicit, implicit, and cross-document types, and a comparative analysis of the corresponding extraction methods and evaluation frameworks is conducted. Findings Academic innovation points exhibit complex and heterogeneous characteristics, including the coexistence of explicit and implicit expressions, multimodal information fusion, and cross-document evolution. Technological development has progressed from rule-based matching to deep learning integrated with semantic modeling, gradually forming a “semantic modeling–knowledge organization–intelligent service” chain. Substantial disciplinary differences are also observed in the structure, representation, and methodological treatment of academic innovation. Research limitations The analysis is based on the available literature and existing innovation point extraction studies, and therefore may not fully capture emerging approaches or rapidly evolving technological developments. In addition, differences in disciplinary contexts and evaluation practices may limit the generalizability of comparisons across research domains. Practical implications The study provides a systematic theoretical and methodological foundation for academic innovation point extraction and offers insights for improving cross-disciplinary knowledge circulation and academic evaluation. The identified extraction paradigms and technology-development trajectory can also inform the design of intelligent and scalable innovation knowledge service systems. Originality/value This study integrates conceptual classification, technological approaches, research themes, and tool ecosystems into a unified review framework for academic innovation point extraction. By distinguishing explicit, implicit, and cross-document innovation points and linking these forms to their corresponding extraction and evaluation approaches, the study provides a structured perspective for understanding the development and future evolution of innovation-oriented scholarly knowledge services.