Empowering Research Through Bibliometric Insights. biblio@vietscies.com Zalo 0916 086 008
Tools

Bibliometrix in R: A Practical Tutorial for Your First Analysis

Run a complete bibliometric analysis with the bibliometrix package: import data, produce descriptive statistics, build networks, thematic maps and three-field plots, or use the biblioshiny interface.

13 min read Updated 2026-11-21 Biblio VietSCI

Bibliometrix is an open-source R package introduced by Aria and Cuccurullo (2017). It covers the full science mapping workflow: importing and converting records, descriptive statistics, citation analysis, networks and thematic maps. Its biggest advantage is reproducibility, because the entire analysis lives in a script you can rerun at any time. This tutorial walks through the core functions you need for a complete first analysis.

Installing R, RStudio and bibliometrix

Step 1: Install R

Download R from CRAN (cran.r-project.org) for your operating system. A recent R release avoids compatibility issues with the current bibliometrix version.

Step 2: Install RStudio

RStudio Desktop, published by Posit, makes it easier to manage projects, inspect data frames and view plots.

Step 3: Install the package

Open RStudio and run the install command in the Console. Downloading the dependencies can take a few minutes.

install.packages("bibliometrix")
library(bibliometrix)
Version note Function names and arguments can change between releases. The code below reflects common usage at the time of writing (November 2026). If something fails, open the documentation with help and the function name, for example help(convert2df).

Importing data with convert2df

convert2df turns a database export into a bibliometrix data frame: one row per document, one column per field tag. Useful tags include AU (authors), TI (title), DE (author keywords), ID (index keywords), CR (cited references), PY (year) and TC (times cited).

# Scopus, CSV export
M <- convert2df(file = "scopus.csv", dbsource = "scopus", format = "csv")

# Web of Science, several plain text files
wosFiles <- c("wos1.txt", "wos2.txt", "wos3.txt")
W <- convert2df(file = wosFiles, dbsource = "wos", format = "plaintext")

# Merge both sources and remove duplicates
combined <- mergeDbSources(M, W, remove.duplicated = TRUE)

Run a quick sanity check with dim(M) to see the number of rows and columns, and table(M$DT) to see the document types. Keep your data files inside the RStudio project folder so paths stay short. For a clean Scopus file, follow our guide on exporting Scopus data.

Descriptive analysis with biblioAnalysis and summary

biblioAnalysis computes the main descriptive indicators: annual output, most productive authors and sources, countries and average citations. summary prints them as tables and plot draws the standard charts.

results <- biblioAnalysis(M, sep = ";")
S <- summary(object = results, k = 10, pause = FALSE)
plot(x = results, k = 10, pause = FALSE)

The argument k sets how many top items are listed. The summary output usually feeds the "Main information about the data" table in a paper: time span, sources, documents, annual growth rate, authors and the share of international co-authorship.

Citation analysis, Lotka and Bradford

Several other functions build on the same data frame. citations(M, field = "article", sep = ";") lists the most cited references in your dataset, and localCitations(M, sep = ";") counts how often each document is cited by other documents in the same collection, which is often more meaningful than global citation counts for a narrowly defined field. lotka(results) tests whether author productivity follows Lotka's law, and bradford(M) divides sources into core and peripheral zones according to Bradford's law. These outputs fill the performance analysis part of a paper and give context for the maps that follow.

Treat these laws as descriptive tools. A good fit tells you the field behaves like many others; a poor fit is not an error, and may simply reflect a young or highly interdisciplinary topic.

Building networks with biblioNetwork and networkPlot

biblioNetwork creates the association matrix and networkPlot draws it. The two key arguments are analysis (the relation) and network (the unit).

analysisnetworkResult
"co-occurrences""keywords", "author.keywords"Keyword co-occurrence network
"collaboration""authors", "universities", "countries"Co-authorship network
"co-citation""references", "sources", "authors"Co-citation network
"coupling""references", "sources", "authors"Bibliographic coupling network
netMatrix <- biblioNetwork(M, analysis = "co-occurrences",
                           network = "author.keywords", sep = ";")

net <- networkPlot(netMatrix, normalize = "association", weighted = TRUE,
                   n = 50, Title = "Keyword co-occurrence", type = "fruchterman",
                   size = TRUE, edgesize = 5, labelsize = 0.7)

For country collaboration you first need to extract country information with one of the package helper functions; check the documentation for your installed version. The argument n caps the number of nodes, much like a threshold in VOSviewer.

Reading the network output

The normalize argument matters more than it looks. Association strength, the same similarity measure VOSviewer uses by default, corrects for the fact that frequent items co-occur with many others simply because they are frequent. Without normalization, the most common keywords crowd the centre of the plot and hide the structure among less frequent but more specific terms. networkPlot also returns cluster memberships, which you can inspect in the returned object and export as a table. Cross-checking these clusters against a VOSviewer map built from the same data is a quick way to see whether your thematic structure is robust or an artefact of one algorithm and layout.

Thematic maps with thematicMap

A thematic map places clusters of keywords on two axes: centrality (how strongly a theme connects to others) and density (how developed it is internally). The four quadrants are read as motor themes, niche themes, emerging or declining themes, and basic themes.

tm <- thematicMap(M, field = "ID", n = 250, minfreq = 5,
                  stemming = FALSE, size = 0.5, repel = TRUE)
plot(tm$map)
Interpret with care Where a theme lands depends on the keyword field, minfreq and your cleaning choices. Do not call a theme "declining" on the strength of a single run; test how stable the picture is when parameters change.

Three-field plots with threeFieldsPlot

A three-field plot is a Sankey diagram linking three dimensions, such as authors, keywords and journals, or countries, institutions and topics. Committees and reviewers find it easy to read.

threeFieldsPlot(M, fields = c("AU", "DE", "SO"), n = c(20, 20, 20))

Keep each field to around 15 to 25 items so the diagram stays legible. The output is interactive, so export a high-resolution static image for the manuscript.

biblioshiny for point-and-click users

If you prefer not to write code, launch the web interface that ships with the package:

library(bibliometrix)
biblioshiny()

A browser app opens where you can load files, filter records and run most analyses with the mouse. It is ideal for exploration, but for a publication you should document every setting you used or move the final steps into an R script so the analysis can be reproduced.

Strengths

No coding, a wide range of ready-made charts, quick export of tables and figures.

Limitations

Harder to keep an audit trail, easy to forget parameters, slower on large datasets.

Recommendation

Explore in biblioshiny, then lock in results with an R script stored alongside your project files.

Pitfalls and practical advice

  • Encoding problems: author names with diacritics break when a CSV is opened and re-saved in a spreadsheet. Always work from the original download.
  • Missing CR column: references were not exported, so co-citation cannot run.
  • Unstandardized keywords: merge synonyms (the synonyms argument in some functions) or drop generic terms (remove.terms), and report the lists you used.
  • No version record: capture R and package versions with sessionInfo() so others can reproduce your work.

A compact script you can adapt

Putting the steps together in one file is the simplest way to make your work reproducible. The script below runs from import to the main figures; adjust file names and parameters to your project.

library(bibliometrix)

M <- convert2df(file = "scopus.csv", dbsource = "scopus", format = "csv")
results <- biblioAnalysis(M, sep = ";")
S <- summary(object = results, k = 15, pause = FALSE)

kwNet <- biblioNetwork(M, analysis = "co-occurrences",
                       network = "author.keywords", sep = ";")
networkPlot(kwNet, normalize = "association", n = 60,
            Title = "Author keywords", type = "fruchterman")

refNet <- biblioNetwork(M, analysis = "co-citation",
                        network = "references", sep = ";")
networkPlot(refNet, n = 40, Title = "Co-citation of references",
            type = "fruchterman")

tm <- thematicMap(M, field = "DE", n = 250, minfreq = 5)
plot(tm$map)

sessionInfo()

Run it top to bottom after any change to the data or parameters, and save the script with the exported figures and tables. If a reviewer asks you to change a threshold or exclude a document type, you rerun the file instead of repeating dozens of clicks. Comment each block with the reason for the settings you chose; those comments become the backbone of your Methods paragraph.

Technical support from Biblio VietSCI

The Advanced and Comprehensive packages of our bibliometric analysis service include bibliometrix outputs (thematic map, three-field plot, trend topics), merged Scopus and Web of Science datasets, and the R script so you can rerun everything yourself. We handle the technical side; interpreting the results and writing the Discussion is up to you. See pricing or submit a request on the order page.

Frequently asked questions

Is bibliometrix free?

Yes. Bibliometrix is an open-source R package installed from CRAN with install.packages. R and RStudio Desktop are free as well.

Can I use bibliometrix without programming?

Yes. Running biblioshiny() opens a web interface for point-and-click analysis. For publication, record your settings or save an R script so the work is reproducible.

Can bibliometrix merge Scopus and Web of Science?

Yes. Import each source with convert2df, then use mergeDbSources with duplicate removal. Check the deduplication result, because the two formats differ.

How do I read a bibliometrix thematic map?

Centrality is on the horizontal axis and density on the vertical axis. Top right holds motor themes, top left niche themes, bottom left emerging or declining themes and bottom right basic themes.

Should I use VOSviewer or bibliometrix?

They complement each other. VOSviewer produces very readable network maps, while bibliometrix is stronger for descriptive statistics, thematic maps and reproducibility. Many studies use both.

Ready to map the knowledge structure of your field?

Send us your research topic. We estimate the dataset size and recommend the right package free of charge within one business day.