Mapping disease-environment connections reported in scientific literature at scale with EasyNER

Climate change and connected environmental disruptions have a large negativeeffect on the health of humans and other species. Improved understanding ofknown links in the environment-disease nexus as well as knowledge gaps isessential to prepare for and mitigate this escalating threat. However, knowledgefragments are scattered across millions of scientific articles making informationsynthesis too complex even for large review consortia. Natural languageprocessing (NLP) tools could help execute this at scale. Here, we developed aninformation synthesis pipeline for mapping links between environmentalprocesses and diseases which expanded our EasyNER tool. We then applied it toextract co-mentions from 16.7 million abstracts in PubMed, a large collection oflife science literature, producing a knowledge graph of the environment-diseasenexus.