Using large language models to help train machine learning SDG classifiers

The Sustainable Development Goals (SDGs) encompass a broad range of interconnected social, economic, and environmental challenges. A vast and rapidly growing body of work aims to advance and evaluate progress toward these 17 goals, to identify ongoing challenges and inform policymaking and related actions. Machine learning methods for automatically classifying text according to the SDGs can help organize and analyze this diverse body of work. However, the multidimensional nature of these goals, combined with their interconnectedness, presents unique challenges for machine learning classification.