ESGformer: Automated multimodal ESG evaluation
Date:
As sustainable development gains momentum, non-financial information has become a critical metric for evaluating enterprises. This shift has led more companies to voluntarily disclose corporate social responsibility (CSR) reports. Most current research on ESG (CSR) reports is conducted by major ESG rating agencies, while academic studies tend to rely on quantitative analyses based on these ratings. However, this approach often overlooks the valuable insights contained within the textual content of ESG (CSR) reports.
We constructed a heterogeneous graph neural network to realize the automatic rating of ESG with multiple styles, and our research results were supported by relevant softwritings, undergraduate scientific research and innovation projects, and undergraduate entrepreneurship programs.