Researchers from the Pacific Northwest National Laboratory have proposed GraphAide, an advanced LLM-based capability that offers insights into domain-specific data and allows users to ask natural language questions.
GraphAide introduces a comprehensive methodology and reference architecture to integrate GenAI with semantic web technologies through a modular and extensible RAG approach.
GraphAide's architecture combines agentic and chain-based approaches to create a complex RAG system that utilizes multiple LLM instances for diverse tasks.
GraphAide demonstrates its effectiveness through the generation of comprehensive KGs from news articles and shows improved NER and relation extraction quality compared to baseline approaches.