Source-linked AI summary
CitNetExplorer: A new software tool for analyzing and visualizing citation networks
Nees Jan van Eck, Ludo Waltman
TL;DR
The paper introduces CitNetExplorer for analyzing and visualizing citation networks, addressing the limited attention given to direct citation networks. It shows how publications build on one another and identifies earlier and later literature connected to a researcher’s work.
Problem
Direct citation networks have received relatively limited attention in scientometric literature, which has focused more on co-networks.
Method
The paper introduces CitNetExplorer and explains concepts including the distinction between essential and non-essential citations for analyzing and visualizing citation networks.
Results
CitNetExplorer makes clear how publications build on each other and can show both earlier literature underlying a researcher’s work and more recent literature influenced by it.
Takeaways & Limitations
CitNetExplorer supports analysis of citation networks and the examination of literature connected across a researcher’s work.
Takeaways & Limitations
The discussion notes that the literature on community detection in networks was not specifically addressed in the prior focus on co-networks.
Abstract
from arXiv · showhide
We present CitNetExplorer, a new software tool for analyzing and visualizing citation networks of scientific publications. CitNetExplorer can for instance be used to study the development of a research field, to delineate the literature on a research topic, and to support literature reviewing. We first introduce the main concepts that need to be understood when working with CitNetExplorer. We then demonstrate CitNetExplorer by using the tool to analyze the scientometric literature and the literature on community detection in networks. Finally, we discuss some technical details on the construction, visualization, and analysis of citation networks in CitNetExplorer.
1. Introduction
CitNetExplorer is introduced as a tool for analyzing and visualizing direct citation networks, addressing comparatively limited attention to this network type. It supports field development studies, research-area delineation, researcher-oeuvre analysis, and literature reviewing.
- Tool and motivation: CitNetExplorer analyzes and visualizes direct citation networks, building on Garfield’s work on algorithmic historiography.Compared with HistCite, it can handle much larger networks, possibly including millions of publications and citation relations.
- Applications: Citation networks can show how a research field develops over time through publications ordered by year and their citation relations.This contrasts with co-citation and bibliographic coupling analyses, which yield a static picture of a field.
- Applications: CitNetExplorer helps delineate research areas by expanding from a core set of relevant publications through citation relations.Users can select publications with at least a specified minimum number of citation relations to the core set.
- Applications: At the micro level, CitNetExplorer maps a researcher’s publication oeuvre to reveal how publications build on one another and whether it contains distinct topic areas.It can also show earlier literature on which the researcher builds and more recent literature influenced by the researcher’s work.
- Applications: CitNetExplorer supports systematic literature search by selecting publications that cite or are cited by a given set of publications.This addresses the limited functionality of bibliographic databases for systematic literature searching.
- Paper scope: The paper introduces CitNetExplorer’s concepts, demonstrates analyses of scientometric and community-detection literature, and discusses technical details of citation-network construction, visualization, and analysis.The tool may also be used for citation networks of patents.
2. Main concepts
CitNetExplorer models citation networks as directed, unweighted, acyclic publication graphs and provides attributes and tools for simplifying, visualizing, and exploring them. Its main concepts include citation scores, transitive reduction, and drill-down or expansion functionality.
- Citation-network structure: Each node represents a publication, while each directed, unweighted edge runs from the citing publication to the cited publication.Citation relations are binary: a publication either cites another publication or does not.
- Citation-network structure: Citation networks must not contain forward-in-time citations or cycles, although real networks can violate these constraints.CitNetExplorer checks for cycles and removes citation relations that cause them.
- Citation-network structure: Acyclic networks allow citations to flow in one direction, support more efficient algorithms, and enable concepts such as transitive reduction.CitNetExplorer uses acyclicity to support visualization, algorithmic efficiency, and analysis concepts unavailable for cyclic networks.
- Publication attributes: Publication attributes include marked, selected, group, complete-record, publication-year, and citation-score information, alongside standard bibliographic data.Marked and selected status support drill-down functionality, while groups and publication years affect visualization.
- Publication attributes: Internal citation scores count citations within the analyzed network, whereas external scores count citations from all publications in the database.CitNetExplorer uses internal citation scores by default.
- Transitive reduction: Transitive reduction removes non-essential citation relations while preserving a path between every publication pair connected by a path originally.CitNetExplorer can display only the relations included in this reduced network, minimizing visual complexity while preserving reachability.
- Drill down and expansion: CitNetExplorer supports drilling down and expanding networks with predecessors, successors, and intermediate publications.These functions are presented as essential for analyzing large citation networks.
3. Demonstration
CitNetExplorer is demonstrated on scientometric literature and community detection, showing how users can inspect, visualize, and analyze large citation networks. The demonstrations cover the user interface, literature selection, network visualization, clustering, and drill-down analysis.
- Applications: The tool is demonstrated on scientometric literature, including the h-index and science mapping, and on community detection in networks.The community-detection application uses almost two million publications and more than 15 million citation relations.
- Scope: The demonstrations are intended to provide a general impression of CitNetExplorer’s possibilities rather than detailed usage instructions.A separate tutorial is available online.
- User interface: The user interface combines network visualization, publication search, bibliographic information, selection controls, and menu-based navigation across subnetworks.The interface is divided into left and right areas, with a menu bar for opening, saving, drilling down, analyzing, and navigating networks.
- Scientometric literature: The scientometric visualizations identify clusters corresponding to science mapping, co-word analysis, social network analysis, and possible interdisciplinarity or an author’s oeuvre.Four clusters are reported, with the green cluster considered difficult to label and potentially representing an author rather than a topic.
4. Technical details
CitNetExplorer constructs citation networks from publication and citation-relation data, validates their temporal and acyclic structure, and positions publications for visualization. It also supports clustering, core-publication identification, and path analysis.
- Network construction: CitNetExplorer can receive publication data alone or publication data together with citation relations.When citation relations are absent, the tool derives them through citation matching.
- Network construction: Citation matching first uses DOI and otherwise requires exact agreement on author, year, volume, and page number.Journal titles are not used because they may be inconsistently recorded.
- Network construction: The constructed network removes citation relations that point forward in time or create cycles.These constraints ensure that the citation network is temporally directed and acyclic.
- Visualization: The visualization places publications in vertical layers representing years and horizontal positions reflecting citation-network closeness.Citations flow upward because citing publications are placed below the cited publications.
- Analysis: The tool offers connected-component extraction, publication clustering, core-publication identification, and shortest or longest path analysis.The first three options treat citations as undirected, whereas path analysis accounts for citation direction.
- Analysis: Clustering assigns closely connected publications to the same group, often allowing clusters to be interpreted as literature topics.The method uses a variant of Newman and Girvan’s modularity function with a resolution parameter.
5. Conclusion
CitNetExplorer is presented as a tool for analyzing and visualizing citation networks, demonstrated through scientometric and community-detection literature. The conclusion positions it as useful for testing and comparing literature-delineation approaches while noting that further experience and development are needed.
- CitNetExplorer analyzes and visualizes citation networks, with the paper covering their construction, visualization, and analysis.
- The tool is demonstrated through analyses of the scientometric literature and literature on community detection in networks.
- The community remains relatively inexperienced with the analyses enabled by CitNetExplorer, so its most useful applications and the strengths of different analysis approaches remain to be learned.
- CitNetExplorer can support literature delineation through either clustering or a combination of keyword search and citation-based expansion.
- The tool can be used to test both literature-delineation approaches and compare their results.
- The authors plan continued development, including possible options related to main path analysis, and hope the tool improves understanding of how research fields develop over time.