Born-reusable scientific knowledge: Concept, implementation, and applications

Abstract

The exponentially increasing growth of scientific literature publication presents a significant challenge to effectively read, process, and fully comprehend the wealth of scientific knowledge. The Open Research Knowledge Graph (ORKG) aims to address this challenge by providing infrastructure that aligns with the FAIR principles, to support the creation, curation, and utilization of scientific knowledge. Nevertheless, the current dependence on crowdsourcing and natural language processing (NLP) for post-publication knowledge extraction restricts the scalability and quality of such knowledge bases. In response to these challenges, we present a novel ’born-reusable’ approach that seeks to create richly-detailed, machine-reusable descriptions of papers directly within the computing environment where the research was conducted, thus placing the onus on authors to ensure their research findings are FAIR prior to publication. With the help of the ORKG R package, salient scientific knowledge is captured from the paper’s associated R source code and serialized to a machine-reusable format (JSON-LD) for harvesting by the ORKG by DOI-lookup. By applying this approach to an unpublished soil science manuscript, we demonstrated how authors are best situated to describe their work in a richlydetailed machine-reusable format. Furthermore, by applying this approach to two published agroecology papers, we demonstrated its relevance to post-publication, thus suggesting that papers which share source code and data sets could be made machine-reusable retrospectively. Finally, a proof-of-concept meta-analysis was conducted to demonstrate how this approach can help facilitate research synthesis by providing FAIR scientific data. We concluded that the ’born-reusable’ approach has promising implications for the reusability of scientific knowledge. However, its broad adoption faces several challenges. Therefore, solutions were explored to improve the approach’s interoperability with knowledge graphs, assist authors with its implementation into their workflows, and strengthen cooperation with publishers to provide the necessary infrastructure.