Retrieval augmented generation (RAG) with enhanced SBERT fine-tuning vector summarization in medical domain
Year: 2026
Type: article
Abstract: An innovative Generative AI-driven medical Chatbot is transforming the healthcare landscape and driving unprecedented advancements in the industry. It provides immediate, convenient, effective, individualized support and compassionate care. Access to accurate and up-to-date healthcare information is a significant challenge, particularly among underserved populations facing disparities in access, language barriers and geographical constraints. To address this issue, a medical chatbot (MediGenius) using the Retrieval Augmented Generation (RAG) architecture implemented through Langchain framework, incorporate vector-based document summarization leveraging advanced Natural Language Processing techniques to deliver accurate and personalized healthcare information. At its core, the Enhanced Siamese Bidirectional Encoder Representations from Transformers (ESBERT) RAG model retrieves relevant information from medical documents. The medical chatbot was rigorously evaluated, and the results demonstrate its effectiveness in delivering accurate healthcare information. Within the RAG architecture, the ESBERT model served as the retriever component. ESBERT outperformed all other retrieval models, achieving the highest precision (93.21%), recall (93.40%), and F1 score (93.89%). These results confirm that the integration of ESBERT into the RAG framework significantly enhances the chatbot’s ability to provide reliable and context–specific medical information.
Source: Expert Systems with Applications
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $3,500
OpenAlex ID: W7150737356