bibliometrix, the R toolkit behind Biblioshiny and the SAAS workflow

Three analysis phases, from raw export to networks

bibliometrix is described as an R-tool for comprehensive science mapping analysis, a package for quantitative research in scientometrics and bibliometrics. The field it serves applies quantitative analysis and statistics to scientific publications and their citation patterns, and the README is clear that this has become essential across scientific fields for evaluating growth, maturity, leading authors, conceptual and intellectual maps, and emerging trends.

The intended users are named: universities, government laboratories, policymakers, research directors, information specialists, librarians and scholars themselves, in the context of research performance evaluation.

The package supports three phases. The first is data importing and conversion to R format from major bibliographic databases. The second is bibliometric analysis of publication datasets, including descriptive statistics, author productivity and source impact. The third is building and visualizing matrices for co-citation, coupling, collaboration and co-word analysis, and those matrices are the input for network analysis, multiple correspondence analysis and other data reduction techniques.

That third phase is what distinguishes the package from a citation counter. A co-citation matrix is not a table you read, it is a graph you analyse, and the choice of matrix is a methodological decision that shapes every conclusion drawn from it.

The SAAS workflow maps analysis onto the structure of a paper

The methodological backbone is the SAAS workflow, a four-stage framework that bibliometrix and biblioshiny are designed around. The point of the framework is transparency and reproducibility, and it works by aligning the analytical pipeline with the structure of a scientific article.

Search covers defining the research question, selecting bibliographic databases, formulating queries and collecting data by web export or API. The named databases are Web of Science, Scopus, OpenAlex, PubMed, Lens.org, Cochrane and Dimensions. Appraisal covers assessing and refining the dataset with a PRISMA flow diagram, applying inclusion and exclusion filters on timespan, language, document type and impact metrics, and ensuring data quality through pre-processing and citation matching.

Analysis performs descriptive analysis at three levels, Sources, Authors and Documents, and then explores the knowledge structures: conceptual through co-word analysis and thematic mapping, intellectual through co-citation and historiograph, and social through collaboration networks. Synthesis generates the visualizations, networks, maps, matrices and Sankey diagrams, and interprets results with the support of Biblio AI, the integrated AI assistant.

The framework is described in a peer-reviewed paper by Aria, Cuccurullo, D'Aniello and Spano in the Journal of Informetrics, which is worth reading before you use the tool, because it is the argument for the way the package is shaped.

Getting bibliometrix and choosing between R and biblioshiny

The package is published on CRAN, and the README badges show the CRAN version badge, a monthly downloads badge from the RStudio mirror and a grand total download counter, alongside an R-CMD-check badge for the build workflow. The package ships with a CITATION.cff file and a NEWS file as well as NEWS.md, and the repository carries an R project file, a DESCRIPTION, a NAMESPACE, R/ and man/ directories, tests/ and a data-raw/ directory, which is the standard CRAN package layout.

The README does not include an installation command block, so the route the documentation points at is the CRAN package page, and the repository badge links there directly. There is also a companion project, biblioshiny, which is bundled with the package and described as bibliometrix for no-coders: a shiny web application providing an intuitive interface for bibliometrix.

That dual entry point is a real design decision rather than a convenience. The same analytical engine is reachable from R for people who want reproducible scripts in a versioned repository, and through a browser interface for people who want the workflow without writing code. The two are the same package, so a study can start in the interface and be reproduced in a script, or the other way round.

The README also notes that the README file itself is generated from README.Rmd, so the source of the documentation is an R Markdown file.

Data management: imports, filters, merging and citation matching

The biblioshiny data management layer is where most of the practical work happens, and its features map onto the appraisal stage of the workflow.

Import and convert handles multiple bibliographic databases: Web of Science, Scopus, PubMed, OpenAlex, Cochrane CDSR, Lens.org and Dimensions. There is API integration for direct retrieval from OpenAlex and PubMed, so a dataset does not have to be exported by hand. Filtering works on publication years, journals, countries, citation counts and custom journal rankings, and collections from different databases can be merged.

Two features stand out for anyone who has done this work manually. The citation matching algorithm reconciles citations across databases, which is where most cross-database projects lose hours. And the PRISMA flow diagram is generated automatically, documenting the selection process from identification through screening, eligibility and final inclusion, which is a reporting requirement rather than a nicety in a systematic review.

The consequence is that a dataset assembled from several sources ends up as one reconciled collection with a documented selection history, which is what makes the downstream analysis defensible.

Three levels of metrics and a logistic growth model

The analytics are organised by the entity being described, and the README names the metrics for each level.

Sources covers journal performance, impact metrics, Bradford's law for concentration of published papers among journals, and source production over time. Authors covers productivity analysis, h-index, Lotka's law for the productivity distribution among authors, collaboration patterns, and author profiles with biographical information pulled from OpenAlex. Documents covers citation analysis, most relevant papers, reference publication year spectroscopy for how a field's attention shifts over time, and trend topics.

There is also a life cycle analysis, which fits a logistic growth model to annual publication counts to identify the growth phase a field is in. That is the kind of analysis a general purpose bibliometric tool will not do for you, and it is the sort of claim that ends up in a research evaluation report, so it deserves the same scrutiny as any other model output.

The visualisation layer includes a main information dashboard summarising the key indicators of the collection, and a three-field plot, a Sankey diagram connecting three metadata fields such as cited references, authors and keywords so that relational patterns are visible at a glance.

What the licence situation and the release record actually say

The licence needs a direct look rather than a summary. The repository carries both a COPYING file and a LICENCE file at the root, and the licence field GitHub shows for the repository is one it cannot classify. That combination usually means the terms are not one of the standard SPDX identifiers, so the answer lives in those files. Anyone planning to redistribute or build a commercial service on top of the package should read them rather than assume a permissive default, and a custom licence is also the kind of thing to check with whoever in your institution handles third-party software terms.

The release record shows steady work. Version 5.4.0 was published on 2026-05-17, 5.4.1 on 2026-06-16 and 5.5.0 on 2026-08-29, with the last push to the master branch on 2026-10-19. The repository is not archived, and the project carries an active status badge, an R-CMD-check workflow badge, a CODE_OF_CONDUCT.md, a CONTRIBUTING.md, a SUPPORT.md and cran-comments.md, which is the full set of files CRAN expects from a package that wants to stay there.

For a package this deeply tied to a published methodology, that maintenance record matters more than it would for a utility. The SAAS framework is the intellectual content, and the code is the implementation of a workflow that people are expected to publish papers about, so version drift changes what a study means when it is re-run two years later.

Editorial conclusion

Adopt bibliometrix if your work needs citation analysis that other people can reproduce, since the SAAS workflow and the published methodology tie each analytical step to a section of a paper, and the package is on CRAN with a documented web interface in biblioshiny. Do not adopt it for one-off citation counting, since the tooling assumes a real study with a documented search, appraisal and inclusion process, and do not assume the AI assistant replaces reading the source data. Before you commit, check that the database you use is among the supported imports, decide whether you will work in R or through biblioshiny, read the licence files yourself because the licence field GitHub shows is one it cannot classify, and note the release record, with 5.5.0 published on 2026-08-29 and the last push on 2026-10-19.

Frequently asked questions

Official sources

1. Issues
2. massimoaria/bibliometrix on GitHub
3. Project website
4. README
5. Releases