KnowUREnvironment: An Automated Knowledge Graph for Climate Change and Environmental Issues

Md Saiful Islam, Adiba Mahbub Proma, Yilin Zhou, Syeda Nahida Akter, Caleb Wohn, Ehsan Hoque

Abstract: Despite climate change being one of the greatest threats to humanity, many people are still in denial or lack motiva- tion for appropriate action. A structured source of knowl- edge can help increase public awareness while also help- ing crucial natural language understanding tasks such as in- formation retrieval, question answering, and recommenda- tion systems. We introduce KnowUREnvironment – a knowl- edge graph for climate change and related environmental is- sues, extracted from the scientific literature. We automatically identify 210,230 domain-specific entities/concepts and en- code how these concepts are interrelated with 411, 860 RDF triples backed up with evidence from the literature, without using any supervision or human intervention. Human evalu- ation shows our extracted triples are syntactically and factu- ally correct (81.69% syntactic correctness and 75.85% preci- sion). The proposed framework can be easily extended to any domain that can benefit from such a knowledge graph.