Content uploaded by Matthias Mueller
Author content
All content in this area was uploaded by Matthias Mueller on Oct 22, 2023
Content may be subject to copyright.
SoftwareX 24 (2023) 101565
Available online 21 October 2023
2352-7110/© 2023 The Author. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
PyblioNet – Software for the creation, visualization and analysis of
bibliometric networks
Matthias Müller
University of Hohenheim, Stuttgart, Germany
ARTICLE INFO
Keywords:
Network
Bibliometrics
Science mapping
Scopus
Python
ABSTRACT
PyblioNet is a software tool for the creation, visualization and analysis of bibliometric networks. It combines a
Python-based data collection tool that accesses the Scopus database with a browser-based visualization and
analysis tool. It allows users to create networks of publication data based on joint authorship, citations, co-
citations, bibliographic coupling, and shared keywords. The overall goal of PyblioNet is to provide valuable
insight and context when conducting research, to help users identify areas for further investigation, and to
support the development of a robust research framework.
Metadata
Nr Code metadata description Please ll in this colum
C1 Current code version V0.8
C2 Permanent link to code/repository used
for this code version
https://github.com/Mat-Mueller/
PyblioNet
C3 Permanent link to reproducible capsule
C4 Legal code license MIT license
C5 Code versioning system used git
C6 Software code languages, tools and
services used
python, html, javascript
C7 Compilation requirements, operating
environments and dependencies
Pybliometrics 3.4.0, NetworkX 2.7.1,
VisJs 9.1.6, Python 3.10
C8 If available, link to developer
documentation/manual
C9 Support email for questions m_mueller@uni-hohenheim.de
1. Motivation and signicance
In recent years, the eld of bibliometrics has gained considerable
momentum in the research community [1]. Within the eld of biblio-
metrics, scientic mapping is a method of visualizing and analyzing
scientic research to identify the structure and dynamics of a eld of
study [2,3]. It uses the information in bibliometric data to determine the
relatedness between publications based on for example shared authors,
direct citation, bibliographic coupling, co-citation, or shared keywords.
The general process of scientic mapping includes the two main
steps of (1) data-collection (and pre-processing) and (2) visualization
and analysis [4,5]. The goal of PyblioNet is to support researchers when
performing bibliometric analyses, scoping reviews or in their daily
literature searches etc. by combining a Scopus data collection and
pre-processing tool with a user-friendly and intuitive analysis and
visualization tool. PyblioNet builds on the Python libraries Pyblio-
metrics [6] and NetworkX [7], as well as the Javascript library vis.js [8],
and can download publication data directly from the Scopus database
via Scopus APIs and visualize/analyze the resulting network using a
browser-based tool.
The three main advantages of PyblioNet in contrast to existing
software capable of using Scopus data (e.g. Bibliometrix [9] Bibexcel
[10], VOSviewer [11], Litstudy [12], CiteSpace [13], or SciMAT [14],
see also [4,15] for detailed overviews of bibliometric software) are: (i)
the ability to work with unique Scopus identiers for authors and doc-
uments, which avoids the need for complex data cleaning and manual
assignment [16,17], (ii) the possibility to collect further information on
citing documents, e.g. to determine co-citation relationships, and (iii)
the ability to include references and citing documents in the analysis and
visualization, e.g. to scope for relevant literature.
More specically, the main functions of PyblioNet include:
- literature searches based on Scopus advanced query strings
- calculation of publication networks based on shared authors, cita-
tion, bibliographic coupling, co-citation and shared keywords
- network manipulation and export of network data
- ltering of network nodes and links (by node type, publication date,
degree centrality, network level or link weight)
- visualization and analysis tools (e.g. searching for specic nodes,
resizing or recoloring nodes, force-directed or hierarchical layout)
E-mail address: m_mueller@uni-hohenheim.de.
Contents lists available at ScienceDirect
SoftwareX
journal homepage: www.elsevier.com/locate/softx
https://doi.org/10.1016/j.softx.2023.101565
Received 31 July 2023; Received in revised form 11 October 2023; Accepted 15 October 2023
SoftwareX 24 (2023) 101565
2
- direct user interaction (e.g. manual repositioning of nodes, node pop-
ups with publication’s abstract, access to the full text of a publication
via its DOI)
In order to use PyblioNet, users need access to the Scopus database
and a valid Scopus API key which needs to be entered upon rst usage.
After that, the user can start entering Scopus advanced search query
strings. After downloading publication data, an HTML le is generated
that can be opened in a browser and contains all necessary data and
visualization/analysis tools.
2. Software description
2.1. CLI data collection
PyblioNet consists of two main components. The rst component is a
Python-based data collection script which is built around the Pyblio-
metrics library [6] to download publication data from the Scopus
database via using Scopus Abstract Retrieval API and Search API
(Fig. 1).
Users can use PyblioNet by executing a Python script, which requires
the installation of the libraries such as Pybliometrics [6], NetworkX [7],
etc. Alternatively, users can run the exe le, which includes all necessary
libraries. For the rst use, users need to enter a valid Scopus API key in
order to access the database via Pybliometrics.
1
After that, users can
start by entering Scopus advanced search query strings.
2
PyblioNet will
display how many publications were found using the search query and
ask the user if they want to continue. If so, the user can continue with a
standard setting, or with an advanced mode where the user can decide
on the following settings:
- Minimum citation count: exclude search results based on their cita-
tions. (default: 0).
- Use cached data if possible: download publication data even if it is
data cached on your computer. (default: yes).
- Download information about citing papers: downloading informa-
tion on publications citing the search results is necessary for co-
citation analysis but takes additional time. (default: yes).
- Create extra nodes for references and citing papers: creating extra
nodes for references and citing papers can result in huge networks
that may be too large to visualize. If the user chooses “later”,
PiblioNet will ask for a minimum occurrence of extra nodes for ref-
erences and citing papers. (default: yes).
- Download abstracts: downloading abstracts for search results in-
creases the size of the html le and takes additional time. (default:
yes).
- Minimum weight for bibliographic coupling: include bibliographic
coupling links between publications only if there are at least x shared
references (reduces network size). (default: 1).
- Minimum weight for co-citation: include co-citation links between
publications only if there are at least x shared citing publications
(reduces network size). (default: 1).
- Minimum weight for shared keywords: include shared keyword links
between publications only if there are at least x shared keywords
(reduces network size). (default: 1).
- Create Gephi le: Creates an additional .gexf le of the network [18]
which can be opened in Gephi. (default: no).
More specically, PyblioNet rst downloads information on the
initial set of publications via Pybliometrics [6], which accesses the
Scopus search API using Scopus advanced search query strings. In a
second step, further information on references of the main search results
are collected using the Abstract Retrieval API (“FULL” and “REF” view).
Finally, citing publications are collected for each main search result via
the Scopus Search API again. The publication data is then used to create
network data using NetworkX [7] where each publication is represented
as a node and relationships between nodes are visualized via links
connecting nodes. PyblioNet covers ve methods to determine network
relationships:
- Shared authorship: nodes are connected if they share one or more
authors (using Scopus author IDs).
- Citation: nodes are connected (via a directed link) if one cites the
other (using Scopus EIDs).
- Bibliographic Coupling: nodes are connected if they share one or
more references (using Scopus EIDs; only for Scopus main results).
- Co-citation: nodes are connected if they share one or more citing
papers (using Scopus EIDs; only for Scopus main results).
- Shared keywords: nodes are connected if they share one or more
keywords (using author keywords; only for Scopus main results).
Fig. 1. UML Activity diagram of the Python data collection tool
1
See also https://pybliometrics.readthedocs.io/en/stable/access.html on
how to access the Scopus database.
2
See also https://www.scopus.com/search/form.uri?display=advanced.
M. Müller
SoftwareX 24 (2023) 101565
3
Finally, a HTML le is created that contains both the network data
and the analysis and visualization tools.
2.2. Browser analysis and visualization tool
The second component is a HTML / JavaScript analysis and visual-
ization tool building on the VisJs [8] package. The visualization and
analysis tool is designed as a browser-based graphical interface that
provides a user-friendly and intuitive way to navigate through complex
publication data.
PyblioNet also allows for different ltering and visualization
methods. First, users can choose to display only publication data from
the main search results or also citing and cited publications. Further
ltering of nodes can be done based on publication date, or the current
degree centrality. Link ltering allows users to easily switch between the
ve network levels authorship, citation, bibliographic coupling, co-
citation or shared keywords [5,19–21]. In the case of bibliographic
coupling, co-citation and keyword relationships, links can additionally
be ltered by their weight (representing the number of commonly cited
or citing literature or keywords).
For visualization, users can enter search queries to highlight nodes
where e.g. the search query “agent based and network” will highlight
nodes that mention “agent based” and “network” in their keywords, title
or abstract. Resizing nodes can be done based on nodes` current degree
centrality (and in case of citation networks also based on their in-degree
or out-degree) or number of citations. Recoloring of nodes can be done
to identify cluster structures within large and dense networks using a
Louvain community detection algorithm [21] as implemented by [22],
or based on common journals (node colors are also used to identify
clusters that are analyzed in more detail using the “Show information”
button). The default visualization of nodes is based on a force-directed
layout algorithm placing well-connected nodes in the center of the
network and less well-connected nodes at the periphery [23]. Users can
also choose to use a hierarchical layout where the y-coordinate of nodes
in the canvas is based on the publication year, hereby positioning older
publications at the top and newer ones at the bottom (in case of a
citation network, for example, the hierarchical layout shows the cu-
mulative nature of a research eld). By changing the spring length users
can dene the distance between nodes and the nodes size denes the size
of the nodes visualized.
Finally, the “Show information” button opens a new window
showing the number of nodes and edges as well as the most frequent
keywords and journals. If users have previously colored nodes (based on
the Louvain algorithm or common journals), additional information for
the communities is displayed. Additionally, users can delete selected
nodes (for selecting multiple nodes, press and hold Ctrl), export the
current set of nodes in a Gephi compatible format [18], or display
additional navigation buttons.
PyblioNet also allows for direct user interaction such as repositioning
nodes manually via drag-and-drop, hovering over nodes to get more
information such as abstract, keywords etc. as well as highlighting nodes
and their direct peers by clicking on a node e.g. to identify related
literature. To quickly access the publication directly from the publisher,
double-clicking on a node opens a new tab using the publication’s DOI
or, if not available, opens google scholar with the publication’s title as a
search query.
3. Illustrative examples
3.1. Literature search
In this section, we illustrate how users can use PyblioNet. A rst use
case of PyblioNet consists of performing literature searches in order to
nd, scan and evaluate relevant academic and scholarly articles or books
to gather information on a specic topic. To illustrate this, we use the
following example search query:
TITLE-ABS-KEY(("innovation diffusion" OR "diffusion of innovation")
AND (agent-based OR multi-agent)) AND DOCTYPE(ar) aiming at nding
relevant published articles within the eld of innovation diffusion and
applying the method of agent-based modelling. The search query results
in 173 search results and in sum, these articles use 6342 references and
are being cited by 1179 publications. After removing duplicates (e.g.
two or more documents use the same reference) a network of 173 main
documents, 4751 references and 967 citing papers is created (see also
Fig. 2).
After opening the network in a browser, users can scan for relevant
literature for example, by starting with well-connected or highly cited
nodes. After identifying an interesting publication, the ve different
network levels allow for scanning nodes’ peers (so called snowballing
[24]) to identify additional relevant publications. As PyblioNet down-
loads and visualizes also cited and citing literature, the snowballing,
however, is not limited to the initial list of results but may contain
further relevant publications not included in the initial search results.
Fig. 2. Example of a citation network of 173 main search results.
M. Müller
SoftwareX 24 (2023) 101565
4
3.2. Science mapping
A second use case addresses research questions within the broad eld
of bibliographic analysis and science mapping. For this we chose the
example of published articles within the journal ‘SoftwareX’, obtained
via the search query “ISSN(23527110)” which resulted in 984 articles
using 24.422 references and 17.961 citing papers. Fig. 3 presents the
resulting networks for the citation, co-citation and keyword analysis as
well as bibliographic coupling.
The citation and co-citation analysis of articles published in Soft-
wareX show only sparsely connected networks where the majority of
nodes remain isolated. This indicates that articles published in the
journal neither recognize related articles nor are considered together by
articles outside the journal. The bibliographic coupling and keyword
analysis, however, indicate a strong common basis and a (perceived)
thematical relatedness. A further analysis of the corresponding key-
words of the communities in the bibliographic coupling network shows
the following information for the two biggest communities.
Cluster: 1. with 177 nodes. Keywords: ‘python’:39, ’machine
learning’:19, ’deep learning’:11, ’image processing’:7, ’gravitational
waves’:6, ’time series’:5, ’optimization’:5, ’computer vision’:5, ’image
analysis’:4, ’data analysis’:4, ’software’:4, ’open-source software’:3,
’tensorow’:3, ’pytorch’:3, ’feature extraction’:3
Cluster: 3. with 138 nodes. Keywords: ‘python’:12, ’openfoam’:7,
’c++’:5, ’visualization’:5, ’computational uid dynamics’:4, ’data
analysis’:4, ’high performance computing’:4, ’high-performance
computing’:4, ’simulation’:4, ’nite element method’:4, ’cfd’:4,
’permeability’:3, ’nite volume method’:3, ’nite elements’:3, ’gpu’:3,
’multiphysics’:3, ’android’:3, ’numerical simulations’:3, ’library’:3,
’turbulence’:3, ’parallel’:3
Finally, analyzing common references connecting articles at the
bibliographic coupling level can be done by switching to a citation
perspective and including references made by the articles. In Fig. 4 we
see the resulting network where important references are highlighted
via scaling the node size based on the in-degree of nodes.
4. Impact
PyblioNet complements the existing set of software designed to help
researchers to perform bibliometric analysis and literature search. It
offers a range of useful features that can change how researchers,
Fig. 3. Example of articles published in SoftwareX.
M. Müller
SoftwareX 24 (2023) 101565
5
educators, or institutions engage with bibliometric analysis, literature
reviews, and knowledge exploration. By combining data collection and
preprocessing with user-friendly visualization and analysis tools,
PyblioNet provides a comprehensive platform for gaining deeper in-
sights into the intricate web of scientic knowledge.
One of PyblioNet’s standout features is its seamless integration with
unique Scopus identiers for authors and documents. This eliminates the
need for time-consuming and error-prone manual data cleaning and
assignment, resulting in more accurate and reliable bibliometric ana-
lyses and allows accessing and incorporating information on citing
documents, and thus, co-citation relationships. The potential use cases
are widespread and range from identifying gaps in existing research,
developing research frameworks, identication of key authors and
works, or synthesizing information.
Requirements for using PyblioNet is a valid Scopus API key as well as
access to the Scopus API. The Python module exists also as standalone
exe le and the visualization and analysis requires only a browser. A
main disadvantage of PyblioNet against other software which uses data
obtained directly from the Scopus homepage is speed and downloading
all information for set of 100 main search results takes up to 15 minutes
and more depending on the internet connections.
5. Conclusions
In this paper, we presented PyblioNet, a software suite designed to
support researchers in bibliometric analysis, scoping reviews, and daily
literature searches. PyblioNet is a valuable addition to the bibliometrics
eld, providing researchers with an efcient and versatile tool for bib-
liometric analysis, literature searches, and science mapping. By
streamlining data collection and offering extensive analysis and visual-
ization capabilities, PyblioNet contributes to a deeper understanding of
scholarly interactions and knowledge dissemination, fostering informed
decision-making and advancing research in diverse domains.
Declaration of Competing Interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Acknowledgments
I gratefully acknowledge the support by the members of the
computational science hub (CSH) of the university of Hohenheim. In
particular, I want to thank Konstantin Kuck, Daniela Bendel and Martin
Müller for their help.
Fig. 4. citation network including information on referenced publications.
M. Müller
SoftwareX 24 (2023) 101565
6
References
[1] Donthu N, Kumar S, Mukherjee D, Pandey N, Lim WM. How to conduct a
bibliometric analysis: an overview and guidelines. J Bus Res 2021;133:285–96.
https://doi.org/10.1016/j.jbusres.2021.04.070.
[2] Small H. Visualizing science by citation mapping. J Am Soc Inf Sci 1999;50(9):
799–813. 10.1002/(SICI)1097-4571(1999)50:9<799AID-ASI9>3.0.CO;2-G.
[3] Chen C. Science mapping: a systematic review of the literature. J Data Inf Sci 2017;
2(2):1–40. https://doi.org/10.1515/jdis-2017-0006.
[4] Moral-Mu˜
noz JA, Herrera-Viedma E, Santisteban-Espejo A, Cobo MJ. Software
tools for conducting bibliometric analysis in science: an up-to-date review. EPI
2020;29(1). https://doi.org/10.3145/epi.2020.ene.03.
[5] Zupic I, ˇ
Cater T. Bibliometric methods in management and organization. Organ Res
Methods 2015;18(3):429–72. https://doi.org/10.1177/1094428114562629.
[6] Rose ME, pybliometrics KJR. Scriptable bibliometrics using a python interface to
Scopus. SoftwareX 2019;10:100263. https://doi.org/10.1016/j.
softx.2019.100263.
[7] Hagberg AA, Schult DA, Swart PJ, Varoquaux G, Vaught T, Millman J. Exploring
network structure, dynamics, and function using NetworkX. editors. In:
Proceedings of the 7th Python in Science Conference; 2008. p. 11–5.
[8] vis.js community. vis.js, 2023. URL: https://visjs.org.
[9] Aria M, Cuccurullo C. bibliometrix An R-tool for comprehensive science mapping
analysis. J Informetr 2017;11(4):959–75. https://doi.org/10.1016/j.
joi.2017.08.007.
[10] Persson O, Danell R, Wiborg-Schneider J. How to use Bibexcel for various types of
bibliometric analysis. Celebrating scholarly communications studies: A festschrift
for Olle Persson at his 60th birthday 2009:9–24. https://portal.research.lu.se/ws/
les/5902071/1458992.pdf.
[11] van Eck NJ, Waltman L. Software survey: VOSviewer, a computer program for
bibliometric mapping. Scientometrics 2010;84(2):523–38. https://doi.org/
10.1007/s11192-009-0146-3.
[12] Heldens S, Sclocco A, Dreuning H, van Werkhoven B, Hijma P, Maassen J, et al.
litstudy: a Python package for literature reviews. SoftwareX 2022;20:101207.
https://doi.org/10.1016/j.softx.2022.101207.
[13] Chen C, CiteSpace II. Detecting and visualizing emerging trends and transient
patterns in scientic literature. J Am Soc Inf Sci Technol 2006;57(3):359–77.
https://doi.org/10.1002/asi.20317.
[14] Cobo MJ, L´
opez-Herrera AG, Herrera-Viedma E, Herrera F. SciMAT: A new science
mapping analysis software tool. J Am Soc Inf Sci Tec 2012;63(8):1609–30. https://
doi.org/10.1002/asi.22688.
[15] Cobo MJ, L´
opez-Herrera AG, Herrera-Viedma E, Herrera F. Science mapping
software tools: Review, analysis, and cooperative study among tools. J Am Soc Inf
Sci Tec 2011;62(7):1382–402. https://doi.org/10.1002/asi.21525.
[16] Strotmann A, Zhao D. Author name disambiguation: What difference does it make
in author-based citation analysis? J Am Soc Inf Sci Tec 2012;63(9):1820–33.
https://doi.org/10.1002/asi.22695.
[17] Baas J, Schotten M, Plume A, Cˆ
ot´
e G, Karimi R. Scopus as a curated, high-quality
bibliometric data source for academic research in quantitative science studies.
Quant Sci Stud 2020;1(1):377–86. https://doi.org/10.1162/qss_a_00019.
[18] Bastian M, Heymann S, Jacomy M. Gephi: an open source software for exploring
and manipulating networks. ICWSM 2009;3(1):361–2. https://doi.org/10.1609/
icwsm.v3i1.13937.
[19] Chang YW, Huang MH, Lin CW. Evolution of research subjects in library and
information science based on keyword, bibliographical coupling, and co-citation
analyses. Scientometrics 2015;105(3):2071–87. https://doi.org/10.1007/s11192-
015-1762-8.
[20] Boyack KW, Klavans R. Co-citation analysis, bibliographic coupling, and direct
citation: Which citation approach represents the research front most accurately?
J Am Soc Inf Sci Tec 2010;61(12):2389–404. https://doi.org/10.1002/asi.21419.
[21] Blondel VD, Guillaume JL, Lambiotte R, Lefebvre E. Fast unfolding of communities
in large networks. J Stat Mech 2008;(10):P10008. https://doi.org/10.1088/1742-
5468/2008/10/P10008. 2008.
[22] Corneliu S., Louvain community detection for Javascript, 2020, [online] Available:
https://github.com/upphiminn/jLouvain.
[23] Jacomy M, Venturini T, Heymann S, Bastian M. ForceAtlas2, a continuous graph
layout algorithm for handy network visualization designed for the Gephi software.
PLoS One 2014;9(6):e98679. https://doi.org/10.1371/journal.pone.0098679.
[24] Wohlin C, Runeson P, Neto PMS, Engstr¨
om E, I CM, de Almeida ES. On the
reliability of mapping studies in software engineering. J Syst Softw 2013;86(10):
2594–610. https://doi.org/10.1016/j.jss.2013.04.076.
M. Müller
