Enhancing Sustainability Reporting Analysis: An AI-Based Approach to Extracting Information from Companies' Reports

Abstract

This thesis centers around the development of an analytical tool for the automated analysis of corporate sustainability reports. Given the increasing attention and importance assigned to companies' ESG issues, an AI solution has been developed for an in-depth analysis of information that a firm decides to disclose. The tool, denominated SearchESG, is part of a process designed to assist companies in the definition of their sustainability strategy. The approach uses OCR technology, NLP techniques for semantic search management, and generative AI-based models customized to return a consistent response. The aim is to help companies kickstart their exploration of relevant sustainability topics for their reports by drawing insights from competitors. Through the SearchESG tool, the company gains a practical starting point based on industry best practices. The tool works on PDF format of ESG reports from competitors that a company wants to analyze. Input from an end user is used through a web application to search and extract related content within the text. The implemented tool arises from several approaches that have been tested to handle the complexity of CSR-related documents, which are distinguished by variety and non-uniformity of content structure. Having to develop a solution that was both specific and generic, the use of artificial intelligence helped address the barrier of the impossibility of standardization.