MedGPT: Enhancing Medical Text Understanding with Hybrid Extractive and Abstractive Summarization and RAG-Powered Q&A.

The rapid growth of medical literature presents significant challenges in terms of efficient information retrieval and comprehension. In this paper, we propose MedGPT, a hybrid framework that enhances the understanding of medical texts through the integration of extractive and abstractive summarization techniques, along with a Retrieval-Augmented Generation (RAG)-powered Question Answering (Q&A) system. The framework utilizes the BigBird-Pegasus model for summarizing medical articles, ensuring that both key information and detailed insights are captured effectively. The hybrid summarization approach combines the strengths of extractive and abstractive methods to produce accurate, concise, and coherent summaries. In addition to summarization, MedGPT incorporates a RAG-based Q&A component, which enables users to pose queries related to the medical texts and receive contextually relevant answers. This interactive system allows for improved accessibility and usability of medical research content. Furthermore, MedGPT is designed for local deployment, addressing concerns related to data privacy and computational efficiency in resource-constrained environments. The proposed system is evaluated based on its ability to summarize lengthy medical documents and generate accurate, context-aware responses to user queries. Experimental results demonstrate that MedGPT significantly enhances the efficiency of medical literature review, supporting timely decision-making and advancing knowledge acquisition in the healthcare domain.