Visualizing the Network of Scholarly Works on DataCite Commons

Collaborative research projects often involve teams of researchers from different institutes, and each research stage—from the planning, the piloting, the data collection, analysis, writing, and revision—produces outputs that embody rich details of the research work that are essential for reproducibility and reuse. Being able to demonstrate and track project outputs simplifies the reporting process for the project leaders, makes visible the contributions of all project members, and supports assessment for funders. Through the FAIR Workflows project, we delivered a project dashboard that showcases the diverse types of outputs produced throughout a project's lifetime. To highlight the diversity of research outputs and their interconnectedness, we designed a visualization of all related works of a project, and implemented it on DataCite Commons for project DOIs as a proof of concept. Now we expanded the scope and launched this network graph for works of all resource types.

The network graph visualizes the connections between different work types. It shows the number of instances of each work type, and hovering over a connection reveals the number of links between any two types. https://commons.datacite.org/doi.org/10.60581/zaev-6p15

Why Connections Matter

One of the key benefits of having openly accessible metadata for a scholarly record is to easily identify and expose the relations among works—for example, investigations conducted in different domains that cite the same literature, results reported by different teams that drew from the same datasets, or projects outputs delivered years ago that inspired new experiments or methodology. These relations have always existed, but are often hidden due to variable archiving conventions and siloed preservation and communication practices. With properly implemented persistent identifier (PID) workflows, however, they can be readily discovered and be fed back into the research process, supporting evidence-based assessment. 

The DataCite Metadata Schema is designed to capture connection metadata for scholarly entities through PIDs and accommodates many different types of relations. When these connections are asserted at the point of sharing —whether creating a data management plan, depositing a dataset, or publishing a preregistration or preprint—they can be recorded in the DOI metadata and incorporated into the PID Graph. Search portals and research tools that integrate with DataCite already exploit the PID Graph to enhance the discoverability of works. With this newly implemented visualization, Commons provides an easy way to demonstrate the reach of these connections, to help researchers and other stakeholders to fully grasp the potential of connection metadata and benefit from it. Here we break down the process behind the creation of the network graph and explore potential avenues for improvement.

Data Generation

The graph uses related identifiers to illustrate the connections between “works” within the system. Related identifiers can be included in the DOI metadata when creating a DOI for a resource (e.g., including the DOI of a related preregistration in the metadata when sharing data), or added to existing DOI metadata over time (e.g., adding the DOI of a dataset to a grant management system when reporting on a funded project). 

To ensure a comprehensive picture of these relationships, the DataCite Commons implementation employs a two-way search method:

1. Incoming Links: First, the system identifies all the "works" that list a given "work" as being "related". These are considered the "incoming links," illustrating the resources that point to the work in question.
2. Outgoing Links: Next, the system examines the "related identifiers" within the primary "work" to pinpoint the "outgoing links," which represent connections the primary “work” has to other resources.

This two-way search creates a more complete pool of "related works" by considering relationships from both perspectives.

Rather than displaying individual works, the network graph groups together works by their resource type. “Resource type” corresponds to the resourceTypeGeneral property of each work in the pool of related works. This focus on resource types allows for a broader visualization of the connections within the system.

Visualization Choices

Visualizing the complex relations between works required many iterations to optimize the clarity of information, the interactivity, and palette consistency with the rest of DataCite Commons. 

• Force-directed layout: The graph employs a force-directed layout, a standard approach for network visualization. This layout uses simulated forces to arrange nodes in a visually appealing and informative manner, allowing users to see the relationships between different resource types and the relative strengths of those relationships.  
• Interactivity: Users can also interact with the graph by dragging and repositioning nodes, providing a dynamic and customizable view.
• Consistent colors: To maintain a unified visual language throughout the system, the network graph uses the same color scheme for resource types as other graphs within the system.

Community Feedback 

We are eager to learn how the community is interacting with the network graph and how it can be further refined to better serve your needs.  By actively incorporating community feedback, the goal is to shape the network graph into an intuitive and powerful tool for exploring and understanding the complex relationships among diverse types of scholarly resources. We invite you to share your ideas and suggestions with us on GitHub or reach out to support@datacite.org.