Facilitating the Analysis of COVID-19 Literature Through a Knowledge Graph

Abstract

At the end of 2019, Chinese authorities alerted the World Health Organization (WHO) of the outbreak of a new strain of the coronavirus, called SARS-CoV-2, which struck humanity by an unprecedented disaster a few months later. In response to this pandemic, a publicly available dataset was released on Kaggle which contained information of over 63,000 papers. In order to facilitate the analysis of this large mass of literature, we have created a knowledge graph based on this dataset. Within this knowledge graph, all information of the original dataset is linked together, which makes it easier to search for relevant information. The knowledge graph is also enriched with additional links to appropriate, already existing external resources. In this paper, we elaborate on the different steps performed to construct such a knowledge graph from structured documents. Moreover, we discuss, on a conceptual level, several possible applications and analyses that can be built on top of this knowledge graph. As such, we aim to provide a resource that allows people to more easily build applications that give more insights into the COVID-19 pandemic.

Knowledge Graphs for COVID-19: An Exploratory Review of the Current Landscape

“…We found four articles that used KGs to facilitate the literature search. In the first paper by Steenwinckel et al [ 5 ], the Kaggle dataset of 63,000 + papers (also known as CORD-19 [ 6 ], released to allow recent advances in natural language processing (NLP) and other AI techniques to generate new insights to fight the pandemic) was used to create a KG. The authors started with a summary of initiatives by other research groups who are using the same dataset, identifying the CovidGraph project as the largest such initiative.…”

“…Four papers were reported under this application group: Steenwinckel et al [ 5 ], Wise et al [ 14 ], Cernile et al [ 18 ], and Michel et al [ 19 ]. All of these papers used the CORD-19 dataset; [ 14 , 18 ] use proprietary code, whereas [ 5 , 19 ] have made their code open-source and followed FAIR principles. While [ 5 , 14 , 19 ] mentioned clustering analysis, there is no way to judge based on these publications which clustering approach is most effective.…”

“…All of these papers used the CORD-19 dataset; [ 14 , 18 ] use proprietary code, whereas [ 5 , 19 ] have made their code open-source and followed FAIR principles. While [ 5 , 14 , 19 ] mentioned clustering analysis, there is no way to judge based on these publications which clustering approach is most effective. To identify influential publications, ref.…”

“…To identify influential publications, ref. [ 5 ] measured node centrality; ref. [ 14 ] measured popularity.…”

“…To identify similar papers, ref. [ 5 ] used nearest neighbors on RDF2vec embeddings, whereas [ 14 ], besides using SciBERT embeddings, also defined novel metrics such as topic and citation similarity. Again, there is no way to assess which approach would be more beneficial to a user.…”