This repository contains the code and data for the paper:
"Impact of EU Non-Financial Reporting Regulation on Spanish Companies’ Environmental Disclosure: A Cutting-Edge Natural Language Processing Approach"
Published in Environmental Sciences Europe
📄 Read the Paper

This study investigates the impact of EU regulations on environmental disclosure quality among Spanish companies from 2015 to 2022. We leverage Natural Language Processing (NLP), specifically ClimateBERT and a fine-tuned version on ClimaText, to analyze 729 corporate sustainability reports.

The study reveals how mandatory disclosure requirements improve specificity and commitment in climate risk reporting, aligning companies with international sustainability frameworks.

climateBERT-analysis/
│── Climate_Disclosure_Analysis.ipynb  # Main Jupyter Notebook
│── data/  # Folder for input reports (not included)
│── results/  # Folder for generated outputs (e.g., graphs, tables)
│── requirements.txt  # Python dependencies
│── LICENSE  # Repository license
│── README.md  # This file
│── Paper.pdf  # Optional: Paper PDF

• 📁 data/ – (Not included) Place corporate reports here before running the notebook.
• 📁 results/ – Stores generated figures and tables from the analysis.
• 📄 Climate_Disclosure_Analysis.ipynb – The core analysis notebook.
• 📄 requirements.txt – Required dependencies for running the notebook.

git clone https://github.com/villacampaporta/climateBERT-analysis.git
cd climateBERT-analysis

Ensure you have Python 3.8+ installed. Then, install the required packages:

pip install -r requirements.txt

• Download corporate sustainability reports from:
  • CNMV Reports
  • FactSet
• Place PDF reports inside the data/ folder.

Start Jupyter and open Climate_Disclosure_Analysis.ipynb:

jupyter notebook

Run the notebook step by step to reproduce the analysis.

• Transformers-based models:
  • ClimateBERT
  • Fine-tuned ClimateBERT using ClimaText
• Text extraction & preprocessing:
  • PyPDF2, pytesseract, wand, Pillow for PDF processing
  • nltk, transformers, scikit-learn for NLP

• Comparing voluntary vs. mandatory disclosure (2015–2022)
• Measuring commitment, specificity, and neutrality in reports
• Assessing alignment with international sustainability standards

📌 How has environmental disclosure changed under regulation?

📌 Model Performance: ClimateBERT vs. Fine-Tuned Model

To replicate this study:

1. Use Python 3.8+ and install requirements.txt
2. Place reports in data/
3. Run Climate_Disclosure_Analysis.ipynb in Jupyter
4. Analyze the outputs in results/

If you use this code or dataset, please cite:

@article{VillacampaPorta2025,
  author    = {Javier Villacampa-Porta and María Coronado-Vaca and Eduardo C. Garrido-Merchán},
  title     = {Impact of EU Non-Financial Reporting Regulation on Spanish Companies’ Environmental Disclosure: A Cutting-Edge Natural Language Processing Approach},
  journal   = {Environmental Sciences Europe},
  year      = {2025},
  doi       = {10.1186/s12302-025-01067-z}
}

This repository is licensed under the MIT License. See LICENSE for details.

Feel free to contribute! Open an issue or pull request if you have suggestions for improvements.

For questions or collaboration inquiries, contact:
📧 [Email] 🔗 [LinkedIn]