Semantic similarity score: A new method for mapping intellectual structures in business research

The rapid growth of scientific knowledge makes traditional science mapping increasingly challenging. Existing methods (e.g., citations, co-citation, and bibliographic coupling) rely heavily on lagging citation data. Consequently, they often fail to construct a comprehensive structural image of scientific fields and are ineffective at tracking novel publications and identifying emerging research. To address these limitations, we propose the Semantic Similarity Score (S3), which leverages transformer models to assess the semantic and conceptual similarity of titles and abstracts. We applied this method across two broad domains, ‘ Innovation and Diversity’ and ‘ Ecosystems in Business’, and evaluated its performance using three performance metrics: Topic Coherence, Silhouette Index, and Quality Index. The results demonstrate that our content-based approach significantly outperforms both established citation-based techniques and standard transformer-based topic modeling methods, offering a more precise approach to mapping both established and emerging intellectual structures.