Biblium: a Python library for comparative bibliometric analysis
摘要
This paper presents Biblium, a Python library for bibliometric and scientometric analysis that introduces a formal framework for comparative group analysis—a capability largely absent from existing bibliometric software. The library’s primary methodological contribution is the BiblioGroup framework, which enables researchers to (1) organize bibliographic datasets into user-defined, potentially overlapping subgroups, (2) perform association testing with standardized residuals to quantify how bibliometric entities distinguish between groups, (3) apply correspondence analysis for dimensionality reduction of group–entity relationships, and (4) classify documents into groups using multiple machine learning algorithms. These capabilities address a fundamental analytical question in comparative scientometrics: what statistically distinguishes one subset of publications from another? Biblium also provides a comprehensive analytical environment supporting six major bibliographic databases (Scopus, Web of Science, OpenAlex, PubMed, Dimensions, Lens.org), descriptive bibliometrics, network analysis, temporal dynamics, and a graphical user interface for researchers without programming expertise. Cross-library consistency testing against five established Python bibliometric libraries demonstrated agreement on all compared metrics. Biblium is openly available on GitHub and PyPI.