Extracting Insights From Climate Change Research with NLP
Climate change has become the biggest threat to our existence. Consequently, research on climate change and mitigation strategies have increased considerably. Scientists and other stakeholders, however, face major challenges in extracting useful information from ever growing scientific literature. Natural language processing (NLP) is a promising approach to analyze large volumes of climate-change related scientific literature. Typically, this approach requires labels for grouping the articles based on user defined criteria. Even labeling a few hundred documents with human subject-matter experts is a time-consuming process. Using climate hazard and risk definitions from the 2021 IPCC report on climate change, we developed a snorkel-based NLP workflow to generate labels for a large climate corpus of 600,000 research articles. We scaled the snorkel labeling across multiple GPUs at the Argonne Leadership Computing Facility. In comparison to months of subject-matter expert labeling process, we labeled the whole corpus in about 12 hours. We used the labeled dataset for a number of downstream tasks to retrieve insights on key climate change trends, risk factors, and environmental impact.
More related videos: • The Future of Data-Centric AI 2022 | ...
More related videos: • AI Data Development: Building Better ...
#climatechangeresearch #naturallanguageprocessing #datalabeling
