Large Language Model-based ESG Disclosure Analysis from Corporate Sustainability Reports
Year: 2026
Type: conference-paper
Abstract: Environmental, Social, and Governance (ESG) disclosures have become an important channel through which corporations communicate their sustainability commitments and governance practices. However, the increasing volume and heterogeneity of corporate sustainability reports make manual analysis inefficient and subjective. To address this challenge, this study proposes a large language model (LLM)-based analytical framework for automated ESG disclosure analysis from corporate sustainability reports. The proposed framework integrates natural language processing techniques with transformer-based large language models to extract, classify, and evaluate ESG-related information from textual disclosures. Specifically, the model identifies key ESG topics and semantic patterns across environmental, social, and governance dimensions, enabling structured assessment of corporate sustainability narratives. Experimental analysis is conducted on a dataset composed of publicly available corporate sustainability reports, where the proposed approach is compared with traditional machine learning and conventional NLP methods. The results demonstrate that the LLM-based framework significantly improves the accuracy and interpretability of ESG information extraction and thematic classification. In addition, the framework provides insights into disclosure transparency and topic distribution across ESG dimensions. The findings contribute to the growing intersection of artificial intelligence and sustainable finance by offering a scalable tool for automated ESG disclosure assessment and supporting data-driven sustainability governance.
Authors: Jinliang Wang, Xiang Ma, Zhaoqiao Ding
Language: English
Subfield: Strategy and Management
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7204168373