COS 40-5 - Transforming knowledge sharing in the field of ecology: A call for machine-actionable scientific information

Abstract: Communicating ecological research efficiently to diverse stakeholders is critical for solving societal challenges like climate change and biodiversity loss. Yet, researchers struggle to share reproducible findings and are confronted with a flood of new publications. While data are now largely pseudo-digitalized as PDF articles, this unstructured format is incompatible with data-oriented support systems such as data filter functions that support automated data extraction. There is a critical need to standardize and fully digitalize the way data is presented in scholarly publications to quickly harvest the data and metadata required for ecological analyses that exploit large datasets.

The Open Research Knowledge Graph (ORKG) is a recent, open-access initiative to address these challenges by providing an interface that enables digitalizing scholarly knowledge in a knowledge graph that is human- and machine-readable. An ORKG knowledge graph can, for instance, provide a condensed overview of a specific research question by comparing scientific results across publications in a tabular format. However, a key barrier to scaling its use is that data published in a PDF format cannot automatically be fed into the ORKG. Rather, it must be manually extracted and entered by scientific experts. Our goal is to address this challenge and motivate a paradigm shift in the way scientific information is shared by creating a tool that allows ecologists to publish their data in a machine-readable format from the outset.

Towards this aim, we developed an ecological case study in the ORKG using a dataset exploring the effect of landscape composition on the incidence of agricultural pests. We analyzed the data in R using a linear mixed effects model and developed an accompanying R script that produces a machine-readable description of the model. The output of this script was saved in a standardized file template, allowing the ORKG to automatically harvest the data and eliminating the need for manual data entry. We are also creating a protocol for adapting the R script to feed output from other common ecological analyses into the ORKG.

Our work has the potential to transform the way we share ecological and other scientific knowledge. Using programming languages like R to facilitate automated data harvesting by the ORKG could feasibly be streamlined into existing FAIR data sharing requirements implemented by many academic publishers. Ensuring data from new publications are automatically stored and structured in the ORKG would make data available for extraction and analysis by scholars in any discipline worldwide.