Abstract
Environmental, social, and governance (ESG) reporting has become a cornerstone of corporate transparency and accountability, especially within high emission sectors such as oil and gas. However, the traditional methods of extracting meaningful insights from ESG data are time-consuming and are in general processed manually.
In this study, the authors introduce a Retrieval-Augmented Generation (RAG) pipeline, which automates the extraction and evaluation of information across large volumes of general and sustainable reporting, enabling analysts to efficiently process and synthesize data from multiple years and companies. The authors propose evaluation metrics that mimick human assessment.
The methodology’s scalability and adaptability make it a promising solution for automating the analysis of corporate ESG disclosures on a large scale, thus providing a robust framework for future research and practical applications in corporate sustainability assessment and climate engagement.
Authors
