Mining RDF Data of COVID-19 Scientific Literature for Interesting Association Rules

In the context of the global effort to study, understand, and fight the new Coronavirus, prompted by the publication of a rich, reusable linked data containing named entities mentioned in the COVID-19 Open Research Dataset, a large corpus of scientific articles related to coronaviruses, we propose a method to discover interesting association rules from an RDF knowledge graph, by combining clustering, community detection, and dimensionality reduction, as well as criteria for filtering the discovered association rules in order to keep only the most interesting rules. Our results demonstrate the effectiveness and scalability of the proposed method and suggest several possible uses of the discovered rules, including (i) curating the knowledge graph by detecting errors, (ii) finding relevant and coherent collections of scientific articles, and (iii) suggesting novel hypotheses to biomedical researchers for further investigation.