Dataset for a Bibliometric Analysis of AI-Enabled E-Assessment in Higher Education (Scopus)

Overview This dataset supports a bibliometric analysis of research on **AI-enabled e-assessment in higher education**. It was compiled to investigate (RQ1) publication and citation trends, (RQ2) the most productive and influential countries, journals, authors, and institutions, (RQ3) the nature and extent of collaboration networks, and (RQ4) the dominant thematic structure and evolution of the field. Data Source and Search Strategy Records were retrieved from the Scopus database (Elsevier) in **July 2026** using a structured query combining three conceptual domains — e-assessment terminology, artificial-intelligence terminology, and higher-education context indicators — linked with the `W/5` proximity operator. The search was restricted to English-language journal articles (`DOCTYPE(ar)`, `SRCTYPE(j)`) published between 2023 and 2026. The complete search string was: ```TITLE-ABS-KEY ( ( ( e-assessment OR "digital assessment" OR "online assessment" OR "formative assessment" OR "summative assessment" OR grading OR feedback OR "automated grading" OR "automated feedback" ) W/5 ( "generative AI" OR ChatGPT OR "large language model*" OR LLM* OR GPT OR "AI-assisted" OR "AI-supported" OR "AI-enabled" ) ) AND ( "higher education" OR universit* OR "tertiary education" ) ) AND DOCTYPE ( ar ) AND SRCTYPE ( j ) AND LANGUAGE ( English ) AND PUBYEAR > 2022 AND PUBYEAR < 2027``` Corpus The initial search returned **295 records**. After duplicate removal (based on Scopus Electronic Identifiers) and title/abstract screening, the final dataset comprises **243 included articles** (52 records excluded with documented reasons). All records are peer-reviewed journal articles in English, published 2023–2026. File Contents Files description:`final_dataset.csv` / `final_dataset.ris` | :The 243 included articles in full Scopus bibliographic format (authors, title, source, year, DOI, abstract, author/index keywords, citation counts, affiliations, open-access status, etc.). |`cleaned_data.csv` / `cleaned_data.ris` | :Cleaned/preprocessed version used for analysis (deduplicated by EID, author names normalized, keywords standardized). | `prisma_screening.csv`: The records screened by title and abstract. `prisma_results.csv` :| Screening decisions (INCLUDE / EXCLUDE) with exclusion reasons. `thesaurus.csv`: Keyword thesaurus mapping semantically equivalent terms to unified preferred forms (e.g., `ai` → `artificial intelligence`). `scopus_export_Jul*.csv` / `*.ris`: The original, unmodified Scopus export files, preserved for reproducibility. Analysis The data were processed with **VOSviewer** and **Bibliometrix/Biblioshiny** for performance analysis and science mapping, including co-authorship analysis (author and country levels), keyword co-occurrence mapping, Bradford's law, Lotka's law, h-index, trend topic analysis, strategic diagrams, and thematic evolution analysis (2023–2024 vs. 2025–2026). Keywords bibliometric analysis; AI-enabled e-assessment; artificial intelligence; higher education; e-assessment; generative AI; large language models; ChatGPT; automated feedback; science mapping; Scopus; VOSviewer; Bibliometrix