Automatic maintenance of COVID-19 related Knowledge Graphs based on large-scale information extraction in scientific literature
Summary
The continuous and ongoing enrichment of publications related to new scientific findings is inspiring, however, it generates new challenges for the scientific community. At any given time, researchers have to evaluate manually a plethora of publications so it is not uncommon for them to spend time reading documents that are irrelevant or inaccurate. The official form of scientific representation is document-based and therefore cannot be processed automatically. Consequently, it is necessary to find innovative ways to process and evaluate scientific text automatically. An approach which has recently been adopted by the scientific community for dealing with the above is the production of Knowledge Graphs (KGs) from scientific text. Aiming to contribute to the necessity for machine-actionable scientific representation, we create the UA-Graph. The UA-Graph is a scientific KG produced in order to assist researchers in finding papers relevant to their interests. Along with the UA-Graph, we present the graph production process. We propose a methodology that processes data and extracts structured scientific information in order to be inserted in the KG. We then design a data model that is general enough to be considered domain-independent and yet precise at the same time. Finally, we implement the data model in a graph DBMS and prove that the produced graph can answer complex queries so as to facilitate scientific research.