Cleaned corpus and reproducible pipeline for the bibliometric analysis of artificial intelligence in civil engineering (2000-2025)

Description

This dataset accompanies the manuscript "Bibliometric Analysis of Artificial Intelligence in Civil Engineering (2000–2025): Trends, Sub-domain Trajectories, Thematic Clusters, and Emerging Research Frontiers". It provides the complete PRISMA-cleaned corpus of 18,334 peer-reviewed Scopus articles, the Python pipeline used for all analyses, the four statistical tests (Pettitt, Mann-Whitney, KS, bootstrap), the 100-record manual audit, the classifier validation, and all bibliometric tables and VOSviewer-ready exports needed to reproduce every reported statistic.

Steps to reproduce

This dataset accompanies the manuscript "Bibliometric Analysis of Artificial Intelligence in Civil Engineering (2000–2025)" submitted to the Journal of Building Engineering. CONTENTS: - corpus_v3_clean.csv: PRISMA 2020-cleaned corpus of 18,334 peer-reviewed journal articles indexed in Scopus between 2000 and 2025, with all Scopus metadata columns preserved. - prisma_log.json: Chiffred PRISMA exclusion log at each of the six filtering steps (kept/excluded counts). - quality_audit_100sample.csv: 100-record manual audit of the rule-based screening (99% precision, Wilson 95% CI 97.0-100%). - classifier_validation.json: Precision/recall/F1 per sub-domain on the 100-record sample (macro F1 = 0.43, weighted F1 = 0.52). - statistical_tests.json: Results of Pettitt changepoint test, Mann-Whitney U tests, Kolmogorov-Smirnov for Lotka law fit, and parametric Poisson bootstrap for CAGR confidence intervals. - indicators.json: Reproducible headline bibliometric indicators. - table_*.csv (11 files): All tabular results referenced in the manuscript. - Python scripts (10 files): Complete pipeline from PRISMA filtering to bibliometric law computation. - VOSviewer maps (3 files): Pre-computed network exports for keyword co-occurrence, co-authorship, and co-citation analyses. USAGE: Run the Python scripts in numerical order (01_prisma_filter.py → 10_vosviewer_export.py) to reproduce every reported statistic. Python 3.11 with pandas, NumPy, and Matplotlib are sufficient; no GPU required. LICENSE: CC BY 4.0 SCOPUS QUERY EXTRACTION DATE: 15 January 2026

Institutions

• Centre Universitaire de Mila

  Mila, Mila

Categories

Licence