[Data & Code] Turning Citation Networks Inside Out: Studying Science Using Content-Based Knowledge Graphs from LLM-Derived Taxonomies

Title:Turning Citation Networks Inside Out: Studying Science Using Content-Based Knowledge Graphs from LLM-Derived TaxonomiesAbstract:Scientific fields are usually studied using citations and other metadata, even though papers primarily transmit knowledge through their content. We present an “inside-out” approach that reconstructs a field structure directly from textual content by converting each paper into a small set of interpretable knowledge components. Using a large language model to induce domain-specific taxonomies and to label papers, we represent every publication as a triplet of measure, data type, and research-question type, and build a knowledge graph whose edges are weighted by shared papers. On a tightly curated corpus of 617 studies of intergenerational wealth mobility, this content-level graph reveals a stable methodological backbone—regression-based mobility measures, alongside substantial temporal variation in how dominant components are recombined into pairs and triads. Beyond frequency, we introduce normalized betweenness-to-connectivity ratios that highlight components and pairings that act as disproportionately important structural bridges relative to their prevalence, pointing to combinations that are central to the knowledge network despite limited uptake. Our work demonstrates how content-derived, taxonomy-driven graphs can complement citation-based science mapping by exposing the evolving architecture of methods, data, and questions that define a research field.Data List: LLM_prompts_summary.py: All the prompts used to generate and assign taxonomy categories. figures.ipynb: To reproduce the figures in the paper with this notebook edge_capa_investigation.ipynb: To investigate the changes in edge capacitance with an additional triplet into the network. (Note that other types of investigations are possible, as functions are added in edge_capacitance_inquiry.py) random_graph.ipynb: To create random graphs. fig_network_graphml_lightcolor.gexf: Gephi file for visualization of the network More can be found in our GitHub repository.