Applying Bibliometrics and a RoBERTa Transformer in the Circular Bioeconomy: A PRISMA 2020 Systematic Review

This exploratory methodological study demonstrates an integrated workflow that combines systematic evidence collection Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA 2020), bibliometric mapping, and Transformer-based natural language processing (RoBERTa) to generate multi-layer insights from Circular Economy-related scholarship, using circular bioeconomy literature as a domain case (2017–2025). Searches across Scopus, ScienceDirect, Taylor & Francis, and SAGE retrieved 2643 records; after deduplication and screening, 50 studies were included (mean quality 13.2/16; 68% high quality). Bibliometric mapping (VOSviewer; Scopus subset n = 1468) revealed three thematic clusters that converge with five conceptual framings extracted via qualitative synthesis, providing cross-method validation of the pipeline’s structural and interpretive outputs. The NLP layer identified a predominantly positive discursive valence in the English-language title–abstract–keyword corpus derived from Scopus records, with declining polarity and increasing subjectivity over time. Because these estimates were obtained from composite bibliographic text fields rather than full-text discussion sections, they should be interpreted as indicators of narrative framing rather than as direct evidence of epistemic bias or empirical overstatement. Within that scope, the joint reading of polarity, subjectivity, hedging, and measurement gaps suggests a possible mismatch between acknowledged evaluative limitations and the caution used to communicate them.