National research signatures: methodological differences in library science ✛
Context
Every scientific paper is like a lab notebook entry, recording not only the results but also the researcher’s handwriting. In library and information science (LIS), methods have long been a subject of introspection: the discipline’s independence hinges on method choice. Look through the lens of countries, and methodological traditions turn out to be strikingly diverse. To unpack these differences, Chengzhi Zhang’s team from Nanjing University of Science and Technology applied Machine Learning to 5281 articles from three key journals published between 1990 and 2019.
Methods
First, a subset of articles was manually annotated with 15 method categories—from bibliometrics and content analysis to surveys and transaction log analysis. Then, a neural network called CogFT, built on SciBERT—a transformer model with an attention mechanism trained to understand the natural language of science—was trained. The model plucks four key text snippets describing the method from the full paper, much like a graphologist picks out the most expressive features of handwriting. Accuracy reached 0,83 macro-F1 after cleaning noisy labels with the Confident Learning algorithm. To compare countries, method frequency vectors were constructed, and distances between them were visualized using UMAP and tested statistically.
Results
The top three—the United States, the United Kingdom, and China—together with another 17 countries account for 90% of the publications. But their methodological recipes differ. In Belgium and the Netherlands, bibliometrics appears in nearly half of all papers—as if their scientists view science through a citation grid. In China and Brazil, experiments dominate, in more than a third of articles. Measuring method diversity with the Shannon index, the same one ecologists use for species richness, we see that nearly all countries have broadened their toolkit over 30 years. Only Belgium remains loyal to a limited set of instruments. Cluster analysis split the trajectories into three groups: “mature” (United States, United Kingdom, Canada)—consistently high diversity; “fast-growing” (China, Australia, and 14 other countries)—a sharp rise; and “traditionalist” (Belgium). The share of articles using multiple methods at once grew from 15% to nearly 50%, with the fast-growing countries driving most of that increase. A comparison on a specific topic, such as “information behavior,” showed that the divergence between national and global profiles, measured by the Jensen–Shannon distance, is shrinking. Scientists are increasingly writing with a shared handwriting, and the differences, confirmed by the Kolmogorov–Smirnov test—a classic of probability theory—are fading.
Implications
The methodological map is no longer a patchwork quilt of local traditions: global co-authorship blurs the boundaries. For LIS itself, this is a chance to maintain its identity among neighboring disciplines. Practically, knowing where expertise lies in experiments versus theory allows for the targeted selection of partners for international projects.
Future development
The next horizon is automatic method classification across languages, to include non-English journals and capture the full picture. Models on the level of Large Language Models will be able to not only recognize but also explain methodological choices, turning scientometrics into a kind of real-time graphology of science. Moreover, the causes of differences need to be explored: economics, research schools, or cultural factors?
Impact
Scientometrics, science policy, bibliometrics—anyone assessing a country’s scientific potential or making collaboration decisions will gain a quantitative tool for comparing methodological profiles.
Next steps
Expand the sample to national journals, use self-supervised learning for weakly labeled data, and conduct surveys of researchers to uncover the motivations behind method choices.
Key open problems
How do you measure the quality of a research method? So far we only assess frequency. It’s like analyzing handwriting without checking whether the text is legible. An open question remains: does the convergence of methodological profiles reflect real integration or just a trend toward the international mainstream?
The knowledge machine's view
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
This study examines methodological patterns in library and information science using 5,281 articles from 81 countries published between 1990 and 2019. Employing machine learning (CogFT based on SciBERT), it classifies 15 methods, builds country profiles, and visualizes distances via UMAP. The analysis reveals persistent national differences, three clusters of methodological diversity trajectories, and a trend toward convergence in specific topics.
Where its strength is
A key strength is the large-scale empirical mapping of methodological traditions using modern NLP tools and rigorous statistical validation. This provides a foundation for comparative analysis of scientific cultures and their evolution.
Where it could go next
- It might be interesting to see how similar approaches to uncovering hidden patterns in large corpora of scientific texts are used in other fields—for example, in genomic data analysis, where neural networks find common 'languages' across different organisms.based on: Genes Speak Different Languages — A Neural Network Found a Common One arXiv:2607.00995
- Another avenue is using AI as a research partner: one could extend CogFT into an interactive tool that not only classifies methods but also suggests unconventional combinations.based on: How Artificial Intelligence Becomes a Scientist's Partner arXiv:2511.20976
- It would be intriguing to compare these results with analyses of the relative effectiveness of general-purpose versus specialized AI models; perhaps a similar question about the 'specialization' of methods across countries could yield new insights from the data.based on: Who Heals Better: General AI or a Narrow Specialist arXiv:2606.28960
Worth trying
- Test whether national methodological differences persist when controlling for research topic and journal type, using the same dataset.
- Apply the trained CogFT model to articles from other disciplines to assess its cross-domain applicability.
- Build a dynamic co-authorship network and examine how its structure relates to the convergence of methodological profiles.
Why this may matter
Understanding how national research traditions form and persist can help design more effective strategies for international scientific collaboration, train early-career researchers, and anticipate the future development of the discipline.
Nearby works in our archive
- Who Heals Better: General AI or a Narrow Specialistsimilarity 0.56
- Cellular Atlases: Who Reads the Language of Tissue Bestsimilarity 0.56
- The Backbone of Food Webs: How to Save Ecosystems in a Changing Climatesimilarity 0.55
- How Artificial Intelligence Becomes a Scientist's Partnersimilarity 0.54
- How Many Mice Does It Take to Understand the Microbiome?similarity 0.54
- Genes Speak Different Languages — A Neural Network Found a Common Onesimilarity 0.53
The archive is sparse around this work: the closest review is only 0.56, against a usual 0.64. It looks like few have drilled nearby.
Original abstract
as written by the authors arXiv:2607.01833
The global development of Library and Information Science (LIS) is influenced by various factors such as the economy, society, culture, discipline, tradition, and more. Consequently, the research methods of LIS vary greatly among countries. To better understand these differences, we conducted a study of 5,281 research papers from 81 countries published in internationally representative journals over the past thirty years. We manually annotated the research methods used in some articles through content analysis, and subsequently developed and trained a deep learning model for automatic classification of research methods. Using this method, we conducted a comparative analysis of the usage of research methods in different countries. Our findings reveal that there are differences in the research methods used across countries, with each country having its unique research profile and distribution of research methods. Even when investigating the same topic, research methods can differ between countries. Our study also uncovers that there are differences between the national and international distribution of research methods, these differences have decreased over the past 30 years. By highlighting the characteristics of discipline development in various countries from the perspective of research methods, our study can help guide discipline development at the national level. This study provides insights into the usage trends of research methods across different countries and highlights the unique characteristics of discipline development in each country. This information can be valuable in promoting collaboration and understanding between countries and in guiding discipline development at the national level.
Belgium is a veritable bibliometrics preserve: 60,58% of its articles use the method, triple the global average. In the US, despite a massive volume of publications, the share of mixed-method papers hasn’t budged in 30 years.
Key numbers
- total number of articles: 5281
- number of countries: 81
- share of U.S. articles: 37.97%
- co-authorship growth: from 1.1 to 1.45 countries per article over 30 years
- share of bibliometrics in Belgium: 60.58%