Mapping knowledge: Topic analysis of science locates researchers in disciplinary landscape

The study presents a new approach for constructing an epistemological coordinate system that locates individual researchers within the disciplinary landscape of science. Drawing on a comprehensive national dataset of scientific outputs, we build a topic model based on a semantic network of publications and terms derived from textual content comprising titles, abstracts, and keywords. Compositional data transformation applied to the topic model enables a geometric analysis of topics across disciplines. The design yields four important results for addressing the gap between knowledge and knowledge-producers. (1) Hierarchical clustering confirms an alignment between traditional disciplinary classification and our empirical, bottom-up topic model. (2) Principal component analysis reveals three axes – Culture–Nature, Life–Non-life, and Materials–Methods – that primarily structure this scientific knowledge space. (3) The projection of individual researchers via their topic portfolios allows to locate them relationally on these three continuous measures of epistemological distinctions. (4) The robustness of our approach is validated by examining the links between researchers’ topic orientation and supplementary variables such as publication practices, gender, institutional affiliations, and funding sources. Our method could inform science policy and evaluation practices, as well as be extended to uncover associations between products and producers in other cultural fields.