[Paper Review] Revealing the research landscape of Master's degrees via bibliometric analyses

Nathalia Chaparro, Sergio Rojas–Galeano|arXiv (Cornell University)|Mar 17, 2021

scientometrics and bibliometrics research55 references5 citations

TL;DR

This paper proposes a bibliometric workflow to analyze the research landscape of Master's degree programs using dissertation metadata, revealing production dynamics, collaborative networks, and thematic structures through text mining and visualization techniques. The approach enables program leaders to assess strengths, identify emerging topics, and guide strategic decision-making, as demonstrated in two engineering Master's programs with reproducible, data-driven insights.

ABSTRACT

The evolution of a Master's programme, like many other human institutions, can be viewed as a self-organising system whose underlying structures and dynamics arise primarily from the interaction of its faculty and students. Identifying these hidden properties may not be a trivial task, due to the complex behaviour implicit in such evolution. Nonetheless, we argue that the programme's body of research production (represented mainly by dissertations) can serve this purpose. Bibliometric analyses of such data can reveal insights about production growth, collaborative networks, and visual mapping of established, niche, and emerging research topics, among other facets. Thus, we propose a bibliometric workflow aimed at discovering the production dynamics, as well as the conceptual, social and intellectual structures developed by the Master's degree, in the interest of guiding decision-makers to better assess the strengths of the programme and to prioritise strategic goals. In addition, we report two case studies to illustrate the realisation of the proposed workflow. We conclude with considerations on the possible application of the approach to other academic research units.

Motivation & Objective

• To develop a systematic, reproducible workflow for analyzing the research output of Master's degree programs using bibliometric techniques.
• To uncover hidden structural and dynamic patterns in academic production, such as collaboration networks, topic evolution, and intellectual influence.
• To support program leaders and evaluators in identifying strengths, weaknesses, and strategic priorities through data-driven insights.
• To demonstrate the applicability of the workflow across different academic units using real-world case studies in engineering and information sciences.
• To provide open-access tools and datasets to enable replication and adoption by other academic institutions.

Proposed method

• The workflow integrates bibliometric techniques including co-occurrence analysis, word frequency extraction, and topic modeling on metadata from Master's dissertations.
• Text preprocessing is applied to fields such as titles, abstracts, keywords, and unigram keywords to standardize and enrich data for analysis.
• Thematic mapping is generated using co-occurrence networks and hierarchical clustering (dendrograms) to visualize conceptual structures.
• Production dynamics are assessed through time-series analysis of annual dissertation counts and citation metrics.
• Interactive visualizations are built using R and Shiny, with a public dashboard and code repository for reproducibility.
• Multiple data fields (e.g., keywords, abstracts, titles) are analyzed separately to assess robustness and consistency of results.

Experimental results

Research questions

• RQ1 How has the research output of a Master's program evolved over time in terms of volume and citation impact?
• RQ2 What are the dominant and emerging research topics within the program’s dissertation corpus?
• RQ3 How do collaborative networks among students and supervisors evolve, and what is the level of institutional or inter-institutional collaboration?
• RQ4 What intellectual and conceptual structures underlie the program’s research output, and how do they shift over time?
• RQ5 To what extent can bibliometric analysis support strategic decision-making in academic program development?

Key findings

• The MSc in Information Sciences (MIS) program produced 170 dissertations between 2012 and 2020, with an average of 18.9 dissertations per year and a collaboration index of 1.43, indicating moderate co-authorship.
• The MSc in Engineering (MIE) program showed a steady increase in annual dissertation output, with 18.9 documents per year on average and 2.09 co-authors per document.
• Thematic analysis revealed that topics in artificial intelligence, software engineering, and geomatics emerged as key research areas in the MIS program after 2011, aligning with program reforms.
• Co-occurrence networks and wordclouds based on keywords and abstracts consistently identified 'machine learning', 'data mining', and 'information systems' as central themes across both case studies.
• The workflow successfully mapped intellectual structures through dendrograms and thematic maps, showing clear clustering of topics such as 'telecommunications' and 'AI' in the MIS dataset.
• The study demonstrated that bibliometric analysis of dissertations can reveal both conceptual and social dynamics, including faculty engagement and topic evolution, with reproducible results via public R scripts and a web dashboard.