TKG-RAG: A Retrieval-Augmented Generation Framework with Text-chunk Knowledge Graph
A retrieval-augmented generation framework that utilizes a text-chunk knowledge graph to improve performance.
The authors propose TKG-RAG, a framework that constructs a text-chunk knowledge graph automatically from domain text. This framework improves Retrieval-Augmented Generation (RAG) performance by addressing limitations such as noise and redundant information in retrieved text chunks. Comparative experiments show that TKG-RAG achieves better accuracy and F1 scores while reducing token consumption.
Based on: TKG-RAG: A Retrieval-Augmented Generation Framework with Text-chunk Knowledge Graph