Dataset and analysis pipeline for: Mapping the emerging field of artificial intelligence in art and design education (2015–2026)
Description
This repository contains the final screened corpus, the full screening decision record, and the open-source analysis pipeline supporting the manuscript "Mapping the Emerging Field of Artificial Intelligence in Art and Design Education: A Bibliometric and Science Mapping Analysis (2015–2026)," submitted to Thinking Skills and Creativity. The dataset comprises 657 journal articles and reviews on AI in art and design education, retrieved from Scopus and the Web of Science Core Collection (11 July 2026), merged, deduplicated, and screened for topical relevance. The screening sheet documents all 1,203 originally identified records and the inclusion/exclusion decision for each, enabling independent verification of the PRISMA-style selection process. The Python pipeline (pandas, NetworkX, matplotlib) performs keyword normalisation, annual production analysis, source and country ranking, keyword co-occurrence network construction with Louvain clustering, Callon thematic mapping, and post-2023 keyword growth analysis, reproducing all tables and figures in the manuscript. Two GraphML network files are included for independent visualisation in VOSviewer, Gephi, or Cytoscape. Raw database exports are not redistributed in line with Scopus/Web of Science usage terms; the derived files here are sufficient for full reproduction.
Files
Steps to reproduce
1. Environment setup Requires Python 3.10+ with pandas, networkx, and matplotlib installed: pip install pandas networkx matplotlib 2. Run the analysis pipeline From the directory containing korpus_v2.csv and bibliometrik_analiz.py, run: python3 bibliometrik_analiz.py korpus_v2.csv --min-kw 5 --split-year 2023 Optional arguments: --min-kw sets the minimum keyword occurrence threshold for the co-occurrence network (manuscript uses 5 for reported statistics; a higher value such as 8 produces the more legible sub-network shown in Figure 2). --split-year sets the year used to split the corpus into pre- and post- generative-AI periods for the keyword growth analysis (manuscript uses 2023). 3. Output The script creates a local "cikti" (output) folder containing: - CSV tables: annual production, top sources, top authors, top countries, most-cited documents, keyword frequencies, thematic map clusters, fastest-rising keywords, country collaboration pairs, Lotka's law distribution - PNG figures (300 dpi): annual production chart, country ranking chart, keyword co-occurrence network, thematic map - GraphML network files importable into VOSviewer, Gephi, or Cytoscape - OZET_RAPOR.md, a plain-text summary of key statistics 4. Verifying the screening process screening_sheet.csv contains the inclusion/exclusion decision for all 1,203 originally identified records (683 from Scopus, 520 from Web of Science Core Collection), retrieved 11 July 2026 using the search string reported in the manuscript's Method section. Cross-referencing this file against korpus_v2.csv (the final 657-record corpus) allows independent verification of the deduplication and scope-screening steps described in the manuscript. 5. Reproducing retrieval from scratch (optional) To retrieve the underlying raw records independently, rerun the Boolean search string reported in Section 3.2 of the manuscript against Scopus and the Web of Science Core Collection, restricting to English-language journal articles and reviews published 2015–2026.