A reproducible bibliometric-content analysis workflow for mapping research fields in computer science

Area of Science:

• Computer Science
• Information Science
• Bibliometrics

Background:

• The exponential growth of computer science literature makes manual synthesis challenging.
• Existing bibliometric methods often lack the interpretive depth for comprehensive field characterization.
• Individual researchers face limitations in synthesizing large volumes of scientific publications.

Purpose of the Study:

• To introduce a reproducible, two-phase workflow integrating bibliometric science mapping and thematic content analysis.
• To enable comprehensive synthesis of computer science literature by individual researchers.
• To identify research gaps and opportunities not apparent through bibliometric analysis alone.

Main Methods:

• A two-phase workflow using R and the bibliometrix package.
• Phase 1: Bibliometric science mapping (keyword co-occurrence clustering).
• Phase 2: Deductive-inductive thematic coding of selected publications with a test-retest reliability procedure (κ = 0.82).

Main Results:

• The workflow was applied to 648 AI-FinTech publications (2017-2026), identifying four thematic clusters.
• Achieved a high inter-rater reliability score (κ = 0.82) without a second coder.
• Uncovered regulatory compliance gaps and AI-blockchain integration opportunities.

Conclusions:

• The integrated workflow provides a feasible and effective method for in-depth literature synthesis by individual researchers.
• The test-retest procedure successfully replaces dual-coder requirements, enhancing accessibility for PhD researchers.
• This approach offers a practical solution for navigating and understanding complex, rapidly growing fields within computer science.