Enhancing Automated Medical Question-Answer Systems Using Fine-Tuned Large Language Models
S. M. Tanvir Uddin, Wasima Mostary, Md. Nazmus Sakib, S. M. Taslim Uddin Raju, Abdul Raqeeb Mohammad · 2025
Automated question and answer (Q&A) systems play a vital role in the medical field by providing accurate information to healthcare professionals and patients. This paper proposes an enhanced approach for automated medical Q&A systems using fine-tuned large language models (LLMs). The MedQuAD dataset is utilized to evaluate the proposed method. In the first stage, various natural language processing (NLP) techniques are applied to clean and preprocess the dataset. This study explored decoder-only models (GPT-2 and Llama2) and encoder-decoder models (Bloom and T5), and fine-tuned them on the MedQuAD dataset. The performance of these models is then compared to determine the most effective LLMs for medical Q&A tasks. The T5 model demonstrates superior performance, achieving a BLEU-4 score of 42.5%, a METEOR score of 36.7%, and a ROUGE-L score of 39.2%, respectively. These results highlight the potential of LLMs to enhance automated medical Q&A systems, significantly improving accuracy and reliability.