Enhancing Library Knowledge Discovery Through Personalized Book Recommendation: An Integrated Analysis

Books remain essential to learning and knowledge access, yet the rapid expansion of digital collections has made effective discovery increasingly difficult. This has heightened the need for personalized book recommendation in digital libraries and discovery platforms. This study presents a library-oriented integrated analysis that combines science mapping techniques with qualitative synthesis to examine the field across four dimensions: methodological trends, public dataset characteristics, benchmark model performance, and future design implications. The findings reveal a clear transition from traditional collaborative filtering to hybrid, graph-based, and deep learning approaches, while persistent challenges such as sparsity, cold-start problems, limited interpretability, and fairness remain unresolved. Widely used public datasets are highly sparse and largely derived from commercial platforms, limiting their suitability for library contexts. Building on these insights, this study further proposes RecSys4Book as a conceptual framework for future library-oriented recommendation design. Overall, the study provides an integrated library-oriented understanding of personalized book recommendation and informs future research and system development in library discovery environments.