Abstract
Throughout the pandemic, research contributing to discoveries associated with COVID-19 has grown considerably. To help increase visibility of this integral body of research and illustrate the extensive organizational collaborations that help move this research forward, a library team utilized data science techniques to analyze COVID-19 research output. These analyses help to demonstrate how libraries can integrate data science expertise to showcase an institutions depth of research engagement and facilitate institutional, national, and global collaboration.
TopBackground
The University of North Carolina at Chapel Hill (UNC-CH) received over one billion dollars in annual research awards in 2021, including over 740 million dollars from federal funding sources. UNC Health Affairs Principal Investigators administered over 800 million dollars (75%) of UNC’s total 2021 research awards, where grants administered by the School of Medicine and Gillings School of Global Public Health (research.unc.edu) were prominent. The UNC School of Medicine is in the 94th percentile among public medical schools for federal grant funding (https://sph.unc.edu/research/). Increasing competition for grant funding and emphasis on demonstrating collaboration among funded researchers provided an opportunity for libraries to envision how skills in DS and data visualization could be leveraged to promote and showcase research outputs and collaborative networks of research teams (Mani et al., 2021).
Bibliometrics is the use of quantitative and statistical analysis to investigate different facets of publication data and is particularly useful for analyzing publication sets too large to effectively review manually (Sugimoto & Lariviere, 2018). Bibliometrics is the scholarly, product-focused component of scientometrics, a “metascience” (Sugimoto & Lariviere, 2018, p. 10) that applies bibliometrics and other quantitative methods to the analysis and measurement of research. Bibliometric indicators (metrics) such as citation counts have long been interpreted as indicators of research impact or influence but not as direct measures of impact (Waltman & Noyons, 2018). Libraries have historically applied bibliometric methods to inform literature retrieval and collection development (Sugimoto & Lariviere, 2018). More recently, libraries are applying these methods and the latest bibliometric analysis and visualization tools (Waltman & Noyons, 2018) to assist their institutions in assessing research output, impact, and the extent of research collaborations (Gutzman et al., 2018). Bibliometric tools, methods, and indicators continue to evolve as scholarly literature, particularly research articles, grows exponentially (Moral-Muñoz et al., 2020), and research institutions, funders, and an increasing diversity of other users strive to quantitatively describe and assess scientific output, impact, and collaboration across disciplines.
Key Terms in this Chapter
VOSviewer: A software tool for constructing and visualizing different kinds of bibliometric networks. The tool has been available as a desktop application since 2021 as an interactive, online application. It is developed and supported by the Centre for Science and Technology Studies (CWTS) at Leiden University.
Altmetrics (Alternative Metrics): Metrics designed to capture attention a resource (e.g., article) is receiving online such as discussions on social media platforms or in blog posts.
Responsible Metrics: The use of quantitative, citation-based measures or indicators (e.g., citation counts, h -index, and journal impact factor) in the evaluation of education and research in ways that are accurate, appropriate, transparent, and ethical.
Citation Analysis: A major method of bibliometrics that examines the quantitative data derived from the use of citations to reference and connect documents. Citation metrics are used to assess the scholarly influence or impact of publications and researchers.
Data Management: The practice of describing and organizing research data to increase understanding and enable further analysis of the data. Well-managed data enables data to be discovered, accessed, and reused to replicate results or repurpose for new research.
Clustering: Type of machine learning that does not require training data or supervision. The k -means algorithm, nonnegative matrix factorization (NMF), and latent Dirichlet allocation (LDA) are examples of clustering algorithms. This is also referred to as unsupervised machine learning .
Scientometrics: The application of quantitative methods such as bibliometrics and citation analysis to the analysis of scientific research.
Data Visualization: The discovery of data insights and communication of findings through techniques that employ the innate ability to distinguish visual patterns.