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

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.本数据集配套论文《2000–2025年土木工程领域人工智能文献计量分析：研究趋势、子领域演进轨迹、主题聚类与新兴研究前沿》("Bibliometric Analysis of Artificial Intelligence in Civil Engineering (2000–2025): Trends, Sub-domain Trajectories, Thematic Clusters, and Emerging Research Frontiers")。本数据集包含经PRISMA标准清洗后的完整文献语料库，涵盖18334篇经过同行评审的Scopus收录论文；提供了用于全部分析的Python分析流水线、四种统计检验方法（Pettitt检验、曼-惠特尼（Mann-Whitney）检验、KS检验、自助法（bootstrap））、100条记录的人工审核结果、分类器验证结果，以及可复现所有报告统计量所需的全部文献计量表格与可直接用于VOSviewer的导出文件。