Graph Augmented retrieval + GenAI = Better RAG in production.
External knowledge is the key to resolving the problems of LLMs such as hallucination and outdated knowledge, which can make LLMs generate more accurate and reliable responses through retrieval-augmented generation (RAG). While revolutionary in capturing semantic meanings, text embeddings often struggle with context sensitivity, contextual meaning, and evolving language use (source paper).
Some of the challenges can be tackled and improved by finetuning a domain-specific embedding model, some would need more advanced retrieval strategies to combine vector search with other search techniques.
Graph RAG is trying to provide a framework and solution components to make RAG more accurate, reliable and explainable by introducing structures of knowledge graph, leveraging graph data science methods and inferencing + generation capabilities of GenAI model.
Github project folder: graph-rag
