Bibliometrics
Bibliometrics (and the closely related scientometrics) is the quantitative study of publications, citations and authorship, used both by researchers studying how science works (the “science of science” or economics of science) and by evaluators of research output. It relies on large citation and abstract databases such as Scopus, Web of Science and Google Scholar, and increasingly on programmatic access to them so that analyses can be scripted, repeated and combined with network, statistical and text-mining tools.
Scriptable access to Scopus with pybliometrics (2019)
Rose and Kitchin present pybliometrics, a Python 3 wrapper for eight Scopus RESTful APIs (document, author and affiliation search and retrieval, plus citation overview and serial title metadata) with consistent class interfaces, on-disk caching and a simple configuration file (Rose 2019). Their worked examples show typical bibliometric tasks: retrieving all 2260 Science papers from 2010 (beyond the 2000-record manual export limit), building a SoftwareX co-authorship network of 734 authors, 228 papers and 1636 edges without name disambiguation thanks to Scopus author IDs, comparing yearly citation distributions with a Kolmogorov-Smirnov test, and topic modeling 226 abstracts with Latent Dirichlet Allocation. Summary: rose2019.
Machine-assisted scoping review of a biomedical literature (2024)
Brown and coworkers combined a formal scoping review with machine learning to map 20 years of literature on nanoparticles targeting tumor associated macrophages (Brown et al. 2024). Searches of five bibliographic databases plus hand searching and Google Scholar gave 28,217 records, 15,374 after deduplication, and 854 after two-stage screening by paired reviewers; a classifier in the Sysrev platform, trained on 7,482 manually screened records, auto-excluded 3,460 low-probability records. Descriptive analysis showed growth inflections the authors tie to the 2011 ipilimumab and 2017 CAR T approvals, a median of about 8 authors per paper, and Biomaterials, Journal of Controlled Release and ACS Nano as the leading journals, before topic modeling of the abstracts. Summary: brown2024.