The idea is:
Implement Graph RAG to improve output quality when conducting document analysis of complex, inter-connected information.
My use case:
Knowledge base with many “context dependencies” like in fields such as finance, scientific research, law, and even coding to some extent, requires the retrieval portion of RAG to populate the context window with higher relevance content, hence resulting in more relevant answers.
I think it would be beneficial to add this because:
RAG ignore context and relationships between documents.
Graph RAG addresses that.
Any resources to support this?
arXiv research article: [2410.05779] LightRAG: Simple and Fast Retrieval-Augmented Generation
Are you willing to work on this?
Let me know how I can contribute.
arXiv suggests using a simpler, much more affordable graph RAG tool:
GitHub - HKUDS/LightRAG: “LightRAG: Simple and Fast Retrieval-Augmented Generation”