An Approach to Medical Response Generation Utilizing GPT-2 Based on Deep Learning

Abstract

 Generative Artificial Intelligence (AI) is transforming healthcare by empowering chatbots to deliver personalized care, accurate diagnoses, and treatment suggestions, which alleviates the workload on healthcare providers. The chatbot system presented here is designed to generate medical responses using Generative Pre-Trained Transformer 2 (GPT-2). Key enhancements include Bidirectional Encoder Representations from Transformers (BERT) for improved question comprehension, Bidirectional and Auto-Regressive Transformers (BART) for summarizing complex inquiries, and a sequence-to-sequence (Seq2Seq) model that ensures relevant response matching. Tested with the COVID Dialogue Dataset, the chatbot achieved a Bilingual Evaluation Understudy (BLEU-4) score of 48.74, demonstrating high-quality responses closely aligned with expert answers. These findings reveal the potential of generative AI to advance telemedicine, providing valuable communication support that can lead to better patient outcomes and more efficient healthcare delivery. The integration of deep learning models in this system enables it to offer responses that are contextually appropriate, accurate, and timely. These results underscore AI's capability to bridge communication gaps between patients and healthcare providers, enhancing patient-provider interactions and contributing to the future development of intelligent healthcare technology.