LukasLiss/scientometric-analysis-africa-affiliated-research

# Scientometric Analysis of African-Affiliated Research about AI in Healthcare Data and analysis code for the scientometric study of machine-learning-for-health research with at least one African-affiliated author, published 2023–2025. Running the pipeline below regenerates every figure, table, and statistic reported in the paper. ## Contents ``` data/ paper_review_process.csv All 4,985 screened records: Paper_ID, first/second reviewer, first/second review decision, followed by the full Dimensions metadata of each paper. included_papers.csv The analysed corpus: the 1,158 papers with an include decision in both review rounds. Papers_step_05_dataset_landscape_llm_cache.csv Cached LLM-extracted dataset names, patient-origin countries, and data modalities per paper. openalex_author_cache.json Cached OpenAlex author-enrichment responses (retrieved June–August 2026). analyses/ 00_full_paper_analysis.py The full analysis pipeline (single entry point). 00_full_paper_analysis.ipynb The same analysis as an executed, cell-by-cell notebook for step-by-step inspection. llm_dataset_extraction.py Provenance: the exact LLM prompt, model, and temperature used to build the dataset-landscape cache (local Ollama, gemma3:12b, temperature 0.1). fig05_institutions_with_countries.py Rebuilds the manuscript's Figure 5 variant of the institution ranking (country names in the labels); run after the main pipeline. requirements.txt Pinned Python dependencies (tested with Python 3.13). ``` ## Reproduce the results ```bash python -m venv .venv && source .venv/bin/activate pip install -r requirements.txt python analyses/00_full_paper_analysis.py ``` This reads `data/included_papers.csv` plus the two caches and writes: - `figures/fig01…fig23_*.png|.pdf` — every figure in the paper, - `figures/tables/*.csv` — machine-readable versions of the paper's tables, - `figures/stats.json` — every scalar statistic cited in the manuscript text. Alternati …