A Retrieval-Augmented Generation Framework for Medical Question Answering: Design, Implementation, and Evaluation of an AI-Driven Healthcare Chatbot

References

[1] Brown, J., Gupta, S.: AI-Driven Chatbots in Healthcare: A Systematic Review. IEEE Transactions on Artificial Intelligence 10(3), 112–126 (2022).

[2] Smith, A., et al.: Machine Learning Applications in Healthcare Chatbots. IEEE Journal of Biomedical Engineering 15(2), 98–110 (2021).

[3] Zhang, P.: Natural Language Processing for Medical Applications. IEEE Transactions on Computational Intelligence 9(5), 305–317 (2020).

[4] Green, L., Adams, T.: Ethical Considerations in AI-Based Healthcare Solutions. IEEE Journal on Ethics in AI 7(1), 56–72 (2023).

[5] Johnson, J., Douze, M., Jégou, H.: Billion-Scale Similarity Search with GPUs. IEEE Transactions on Big Data 7(3), 535–547 (2021).

[6] Thomas, M.: Comparative Study of LLMs in Healthcare AI. IEEE Computational Healthcare Journal 20(3), 112–129 (2021).

[7] Kim, H., Lee, W.: Improving Medical Chatbot Responses with LangChain Framework. IEEE Transactions on Medical Informatics 5(2), 90–105 (2022).

[8] Patel, D., Ramesh, K.: Security and Privacy in AI-Based Healthcare Assistants. IEEE Security and Privacy Journal 12(5), 188–202 (2023).

[9] Lewis, P., et al.: Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. In: Advances in Neural Information Processing Systems (NeurIPS), pp. 9459–9474 (2020).

[10] Chen, B., et al.: Impact of AI Chatbots on Patient Engagement. IEEE Healthcare Informatics Review 14(2), 75–89 (2021).

[11] Kumar, S.: Real-Time Query Processing in AI Healthcare Assistants. IEEE Transactions on Computational Intelligence 8(3), 200–215 (2023).

[12] Davies, L.: Challenges in AI-Assisted Medical Consultations. IEEE Journal of Medical AI Ethics 10(4), 160–175 (2022).

[13] Li, K., Wang, H.: Multi-Modal AI for Healthcare Chatbots. IEEE Transactions on AI Robotics 17(2), 320–335 (2021).

[14] Singh, A.: Personalisation in AI-Based Healthcare Chatbots. IEEE Medical Intelligence Journal 9(3), 220–234 (2023).

[15] Reimers, N., Gurevych, I.: Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. In: Proceedings of EMNLP, pp. 3982–3992 (2019).

[16] Jackson, R., Ford, B.: Scalability of AI-Driven Medical Assistants. IEEE AI Applications Review 13(2), 65–80 (2022).

[17] Roberts, P.: Natural Language Understanding in Medical Chatbots. IEEE Computational Linguistics Journal 11(4), 112–128 (2023).

[18] Martinez, C., et al.: Bias and Fairness in AI-Based Healthcare Systems. IEEE AI Ethics Journal 15(3), 99–115 (2021).

[19] Nelson, J.: Improving AI Chatbot Performance with Adaptive Learning. IEEE Transactions on Adaptive AI 16(5), 205–220 (2022).

[20] White, T.: Hybrid AI Models for Healthcare Assistance. IEEE Computational Biology Journal 7(1), 45–60 (2023).

@article{1717085,
      author = {Swarnava Banerjee, Anindita Bhattacharjee, Apurba Paul},
      title = {A Retrieval-Augmented Generation Framework for Medical Question Answering: Design, Implementation, and Evaluation of an AI-Driven Healthcare Chatbot},
      journal = {Iconic Research And Engineering Journals},
      year = {2026},
      volume = {9},
      number = {10},
      pages = {3632-3638},
      issn = {2456-8880},
      url = {https://www.irejournals.com/formatedpaper/1717085.pdf},
      abstract = {The use of artificial intelligence systems in delivering evidence-based medical knowledge to patients is one promising avenue. Despite their popularity, existing large language models (LLMs) tend to suffer from hallucinations and outdated knowledge, which poses serious risks when such tools are used in the sensitive healthcare industry. In this work, we explore a retrieval-augmented generation (RAG) architecture combining FAISS-based dense vector retrieval with the Llama-3.3-70B language model, coordinated through the LangChain framework, to overcome these disadvantages. Specifically, our method involves the encoding of a selected collection of medical literature into the 384-dimensional dense vector space using the all-MiniLM-L6-v2 sentence transformer model. These vectors get stored in a FAISS flat-L2 store, after which they are retrieved during inference and used by the LLM as context when generating answers in order to improve their relevance and accuracy. We also apply a constraint on the type of prompts given to the LLM in order to restrict the responses to only medical information. Evaluation done qualitatively and quantitatively confirms the effectiveness of our approach in providing high-quality and reliable medical.},
      keywords = {Retrieval-Augmented Generation (RAG), Large Language Models, Healthcare Chatbot, FAISS, LangChain, Natural Language Processing, Medical AI},
      month = {April},
      doi = {https://doi.org/10.64388/IREV9I10-1717085}
  }