Science Mapping AI-enabled Digital Transformation in Project Management through a Combined Bibliometric and BERTopic Modelling Approach

Abstract

This repository contains the analytical datasets, validation outputs, sensitivity-analysis materials and supporting documentation associated with the study: Science Mapping AI-enabled Digital Transformation in Project Management through a Combined Bibliometric and BERTopic Modelling Approach The study combines bibliometric science mapping and BERTopic-based semantic topic modelling to examine the explicit and latent knowledge structures of AI-enabled digital transformation research in project management. The final analytical corpus comprises 562 peer-reviewed journal and review articles published between 2017 and June 2026. The repository includes: a cleaned dataset prepared for BERTopic analysis; a bibliometric dataset prepared for VOSviewer analysis; archived outputs from five BERTopic sensitivity configurations; topic-coherence, NPMI, silhouette-score, topic-count and outlier-count results; screenshots documenting alternative topic solutions; manuscript figures and, where available, their source data; a README file describing the files, analytical settings and data restrictions. The final BERTopic configuration identified 16 substantive topics and retained 94 documents as Topic −1 outliers. The reported validation values for the final model were c_v = 0.433, NPMI = −0.126 and silhouette score = 0.048. The sensitivity analysis varied the UMAP n_neighbors parameter and the HDBSCAN min_cluster_size parameter. The archived outputs are provided to support transparency regarding model selection and the parameter dependence of smaller topic boundaries. The repository supports verification and transparency of the reported science-mapping and semantic-analysis outputs. It does not contain the complete executable analysis notebook or serialised alternative BERTopic models. The main modelling procedures, parameter settings, validation methods and sensitivity configurations are documented in the associated article and supplementary materials. Some bibliographic metadata and abstracts were obtained from licensed databases, including Scopus, Web of Science, IEEE Xplore and Google Scholar. Redistribution is therefore limited to the extent permitted by the relevant providers. Where full records cannot be shared, the repository provides cleaned or derived analytical data and aggregate outputs.