COVIDPUBGRAPH: A FAIR Knowledge Graph of COVID-19 Publications

Area of Science:

• Biomedical Informatics
• Computational Biology
• Information Science

Background:

• The exponential growth of scientific literature on SARS-CoV-2 and COVID-19 presents challenges for researchers seeking comprehensive insights.
• Effective data aggregation and access are crucial for advancing research in the ongoing pandemic.

Purpose of the Study:

• To develop and present COVIDPUBGRAPH, a Linked Data and FAIR principles-compliant RDF knowledge graph.
• To facilitate rapid access to a curated, comprehensive dataset of COVID-19 scientific publications for researchers.

Main Methods:

• Utilized the CORD-19 dataset as the primary data source.
• Applied advanced natural language processing techniques including named entity recognition, entity linking, and link discovery.
• Constructed an RDF knowledge graph, ensuring regular bi-weekly updates.

Main Results:

• The current COVIDPUBGRAPH contains over 268 million triples.
• The knowledge graph is interconnected with 9 external datasets via more than 1 million links.
• Use case studies demonstrated the utility of COVIDPUBGRAPH for diverse research applications.

Conclusions:

• COVIDPUBGRAPH provides a valuable, interconnected resource for researchers navigating COVID-19 scientific literature.
• The knowledge graph adheres to Linked Data and FAIR principles, promoting data interoperability and accessibility.
• The resource is publicly available, fostering collaborative research and knowledge discovery in the field of COVID-19.