Methodological Approaches to ESG Disclosure Analysis: A Systematic Review of the new Machine Learning Techniques

This thesis explores recent methodological approaches to Environmental, Social, and Governance (ESG) disclosure analysis, with a focus on how machine learning (ML) and natural language processing (NLP) techniques are being used to address the challenges of unstructured and fragmented ESG data. As ESG reporting becomes increasingly important for investors, regulators, and other stakeholders, there is a growing need for tools that can handle complex, large-scale information in a reliable and interpretable way. A key part of this research involves analyzing different ML techniques, such as Support Vector Machines, Random Forest, XGBoost, BERT-based models like FinBERT, and multimodal deep learning approaches. For each, the study highlights the main advantages and disadvantages, and compares their performance in terms of accuracy, interpretability, and scalability. The goal is to identify which method is best suited for ESG disclosure analysis, and to provide insights that can support both academic research and practical applications in the field.