Autors: Botev, A. N., Lazarova, M. K., Nakov, O. N.
Title: Rag-Based Framework for Intelligent Banking Assistants Leveraging Structured Financial Data
Keywords: enterprise knowledge management, intelligent banking assistants, regulatory compliance, retrieval-augmented generation, structured data

Abstract: The paper presents a novel Retrieval-Augmented Generation (RAG) framework for intelligent banking assistants, integrating structured financial and regulatory data to improve accuracy and compliance in enterprise scenarios. A two-stage architecture RAG system is suggested. At the embedding stage advanced semantic chunking and metadata enrichment of the documents is utilized. At the retrieval and generation stage, a single input is used merging the original prompt, retrieved context, and generation instructions, and is passed to an LLM to generate the final, compliance-checked response. The suggested methodology is experimentally evaluated with structured Excel QA pairs and unstructured PDF policy documents. Manual expert evaluation confirm the effectiveness of combining RAG with structured data for complex banking knowledge tasks, highlighting practical considerations for real-world deployment.

References

  1. P. Lewis, E. Perez, A. Piktus, F. Petroni, V. Karpukhin, N. Goyal, H. Küttler, M. Lewis, W. Yih, T. Rocktäschel, S. Riedel, and D. Kiela, “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks,” in Advances in Neural Information Processing Systems, vol. 33, 2020, pp. 9459-9474.
  2. V. Karpukhin, B. Oguz, S. Min, P. Lewis, L. Wu, S. Edunov, D. Chen, and W. Yih, “Dense Passage Retrieval for Open-Domain Question Answering,” in Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2020, pp. 6769-6781.
  3. A. Shuster, S. Poff, M. Ung, J. Weston, and A. Szlam, “Retrieval Augmentation for Creative Text Generation,” arXiv preprint arXiv:2102.06405, 2021.
  4. F. Petroni, A. Piktus, A. H. Miller, P. Lewis, M. Lewis, V. Karpukhin, S. Riedel, and W. Yih, "KILT: a benchmark for knowledge intensive language tasks," in Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2021, pp. 2523-2544.
  5. H. van den Broek-Altenburg, A. S. Ather, and A. L. Williams, “The potential of large language models for healthcare: a review of the applications and challenges,” Journal of the American Medical Informatics Association, vol. 30, no. 11, pp. 2005–2016, 2023.
  6. P. Shah, A. Chintalapati, and R. Srivastava, "FinGPT: Instruction Tuning for Financial Large Language Models," arXiv preprint arXiv:2310.04791, 2023.
  7. H. Yu, "Legal NLP: A review of the state-of-the-art," Natural Language Engineering, vol. 29, no. 2, pp. 235-261, 2023.
  8. W. Eisenschlos, S. Thorne, F. Petroni, and S. Riedel, "Fusing Structure and Text for Richer Representations," in Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, 2021, pp. 3206–3220.
  9. P. Sharma, and P. Singh, "A Comprehensive Review on Conversational AI in the Banking Sector," Journal of King Saud University - Computer and Information Sciences, vol. 34, no. 10, pp. 8239-8255, 2022.
  10. A. F. Adadi, and M. Berrada, "Explainable AI for Banking and Finance: A Systematic Literature Review," IEEE Access, vol. 8, pp. 169336-169357, 2020.
  11. I. Koychev, P. Nakov, P. Darakchiev, and P. Stanchev, “bgGPT: A Bulgarian Generative Pre-trained Transformer,” in Proceedings of the International Conference on Recent Advancesin Natural Language Processing (RANLP 2023), 2023, pp. 627–635.

Issue

2025 13th International Scientific Conference on Computer Science, COMSCI 2025 - Proceedings, 2025, Albania, https://doi.org/10.1109/COMSCI67172.2025.11225090

Вид: публикация в международен форум, публикация в реферирано издание, индексирана в Scopus