Knowledge Graph-Driven Retrieval-Augmented Generation: Integrating Deepseek-R1 with Weaviate for Advanced Chatbot Applications
Alexandru Lecu, Adrian Groza, Lezan Hawizy
TL;DR
The paper tackles the reliability gap in LLM-based biomedical chatbots by grounding generation in a curated knowledge graph for age-related macular degeneration (AMD). It presents an integrated architecture combining a GraphDB-backed ontology, Weaviate-powered embedding search, and a locally deployed Deepseek-R1:7B model to enable retrieval-augmented generation with verifiable provenance. A dedicated CausalAMD ontology, an end-to-end pipeline for extracting and refining causal relations from abstracts, and a robust RAG workflow with provenance are introduced. Experimental and architectural results demonstrate reduced hallucinations, improved factual accuracy, and clearer, evidence-backed responses, highlighting practical utility for AMD-focused chatbot applications and potential generalization to other biomedical domains.
Abstract
Large language models (LLMs) have significantly advanced the field of natural language generation. However, they frequently generate unverified outputs, which compromises their reliability in critical applications. In this study, we propose an innovative framework that combines structured biomedical knowledge with LLMs through a retrieval-augmented generation technique. Our system develops a thorough knowledge graph by identifying and refining causal relationships and named entities from medical abstracts related to age-related macular degeneration (AMD). Using a vector-based retrieval process and a locally deployed language model, our framework produces responses that are both contextually relevant and verifiable, with direct references to clinical evidence. Experimental results show that this method notably decreases hallucinations, enhances factual precision, and improves the clarity of generated responses, providing a robust solution for advanced biomedical chatbot applications.