Medical Assistant Utilizing Large Language Models with Retrieval Augmented Generation and Vector Search
Abstract
A trending strategy in the development of domain-specific assistants is to integrate the concepts of the Retrieval-Augmented Generation (RAG) with Large Language Models (LLMs) to build efficient high-quality assistants. The given research will develop a chatbot medical assistant offering credible data on health conditions, diseases, treatments, and preventative measures. The knowledge base is a pre-processed knowledge base based on “The Gale Encyclopedia of Medicine (Second Edition).” A custom set of domain-specific embeddings is produced with the PubMedBERT model and embedded representations are stored in the Qdrant cloud vector database, serving to allow rapid retrieval. The use of cosine similarity means that it uses the vector search which is provided to a RAG pipeline supported by two LLMs, LLaMA-3 70B through Groq API and Google MedGemma 4B through Hugging Face API. I compared the PubMedQA datasets to conduct an evaluation on their performance with metrics being ROUGE-L, BLEU scores, and response latency. The preliminary results show that the LLaMA-3 with the help of PubMedBERT embeds is considerably ahead of MedGemma, reaching the ROUGE-L score of 0.1741 and BLEU score of 0.0229 with lower latency. The chatbot was implemented on Streamlit framework which supports session-based memory to create continuity in conversation. The proposed research proves that the combination of LLMs, vector search, and RAG pipelines is effective to create medicalspecific assistants.
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.