[Paper Review] Advanced Unstructured Data Processing for ESG Reports: A Methodology for Structured Transformation and Enhanced Analysis

Jiahui Peng, Jing Gao+7 more|arXiv (Cornell University)|Jan 4, 2024

Sustainable Industrial Ecology4 citations

TL;DR

This paper proposes a novel methodology using the 'Unstructured Core Library' to transform unstructured ESG reports into structured, analyzable formats through high-precision text cleaning, image-based text extraction, and table standardization. The approach enables accurate handling of diverse layouts and data types across industries, significantly advancing NLP and LLM applications in corporate sustainability assessment.

ABSTRACT

In the evolving field of corporate sustainability, analyzing unstructured Environmental, Social, and Governance (ESG) reports is a complex challenge due to their varied formats and intricate content. This study introduces an innovative methodology utilizing the "Unstructured Core Library", specifically tailored to address these challenges by transforming ESG reports into structured, analyzable formats. Our approach significantly advances the existing research by offering high-precision text cleaning, adept identification and extraction of text from images, and standardization of tables within these reports. Emphasizing its capability to handle diverse data types, including text, images, and tables, the method adeptly manages the nuances of differing page layouts and report styles across industries. This research marks a substantial contribution to the fields of industrial ecology and corporate sustainability assessment, paving the way for the application of advanced NLP technologies and large language models in the analysis of corporate governance and sustainability. Our code is available at https://github.com/linancn/TianGong-AI-Unstructure.git.

Motivation & Objective

• To address the challenge of analyzing unstructured ESG reports with inconsistent formats and complex content.
• To develop a robust framework capable of processing diverse data types—text, images, and tables—across varying report layouts and industry standards.
• To enable high-precision transformation of ESG reports into structured formats for improved downstream analysis using NLP and large language models.
• To support industrial ecology and corporate sustainability assessment by standardizing unstructured ESG data for scalable analysis.

Proposed method

• The methodology employs the 'Unstructured Core Library' as the central processing framework for ESG report data.
• It applies high-precision text cleaning to remove noise and standardize textual content from ESG reports.
• It uses advanced computer vision and OCR techniques to extract and interpret text from images within reports.
• It implements table normalization to standardize tabular data across different report structures and layouts.
• The approach supports multi-modal data processing, handling text, images, and tables in a unified pipeline.
• It integrates with NLP and large language model (LLM) workflows to enable enhanced analysis of transformed ESG data.

Experimental results

Research questions

• RQ1 How can unstructured ESG reports with diverse formats and layouts be systematically transformed into structured, analyzable data?
• RQ2 What techniques enable high-precision extraction of text from images and tables in ESG reports?
• RQ3 To what extent can a unified methodology handle variations in report structure across different industries?
• RQ4 How does the proposed method improve the accuracy and consistency of ESG data for downstream NLP and LLM-based analysis?
• RQ5 What is the impact of structured ESG data on industrial ecology and corporate sustainability assessment?

Key findings

• The methodology successfully transforms unstructured ESG reports into standardized, machine-readable formats with high precision.
• Text extraction from images achieves high accuracy through advanced OCR and layout-aware processing.
• Table normalization effectively standardizes tabular data across heterogeneous report structures.
• The approach enables consistent data representation across diverse industries and reporting styles.
• The transformation pipeline enhances the reliability and scalability of ESG data for NLP and LLM-based analysis.
• The method provides a foundation for large-scale, automated sustainability assessment using structured ESG data.