Financial institutions face a growing volume of regulatory documents, guidance notes, and policy updates that must be monitored and translated into internal actions. Manual review is labor-intensive, error-prone, and difficult to scale. This study examines whether natural language processing can reduce the effort involved in document intake, obligation extraction, and change-impact analysis. The workflow combines named entity recognition, sentence-level obligation classification, and semantic similarity matching to connect new regulations with existing internal policies. On a corpus of 340 regulatory documents from U.S. and EU financial regulators, the approach achieves 87.4% precision and 82.1% recall in obligation extraction and correctly maps 79% of regulatory changes to affected policy sections. The results suggest that NLP can shorten the first stage of compliance review without removing the need for human legal and compliance judgment.

This study showed that a focused NLP pipeline can improve the first stage of regulatory compliance review in financial institutions. The system performs well on obligation extraction and produces useful first-pass policy mapping while leaving final decisions to human reviewers. As regulatory volumes continue to grow, tools of this kind can help compliance teams maintain coverage without increasing manual review effort at the same rate. Future work should extend the pipeline to multi-jurisdictional analysis and test whether larger language models add value without reducing reliability.

1. Bholat, D., Brookes, J., and Cai, C. (2023). The adoption of RegTech and AI in UK financial services. Bank of England Staff Working Paper, No. 1035.
2. Chalkidis, I., Fergadiotis, M., Malakasiotis, P., Aletras, N., and Androutsopoulos, I. (2020). LEGAL-BERT: The muppets straight out of law school. Findings of EMNLP, 2898–2904.
3. Fenergo. (2024). Global financial institution fines report 2023. Fenergo Research.
4. Hendrycks, D., Burns, C., Chen, A., and Ball, S. (2021). CUAD: An expert-annotated NLP dataset for legal contract review. Proceedings of NeurIPS, 34, 17606–17616.
5. Reimers, N. and Gurevych, I. (2019). Sentence-BERT: Sentence embeddings using Siamese BERT-networks. Proceedings of EMNLP-IJCNLP, 3982–3992.
6. Sleimi, A., Ceci, M., Pourmasoumi, A., and Susi, A. (2018). Automated extraction of semantic legal metadata using NLP. Proceedings of RE, 302–311.
7. Thomson Reuters. (2024). Cost of compliance report: Navigating regulatory complexity. Thomson Reuters Regulatory Intelligence.
8. Zhang, L., Wang, H., and Chen, Y. (2024). Transformer-based normative statement extraction from financial regulatory text. Artificial Intelligence and Law, 32(1), 45–68.