A KBQA Method Based on Mixed Retrieval Augmented Generation Technology

Retrieval augmented generation technology has made it significantly convenient to utilize large language models for knowledge base question answering on local document data, and it has become one of the prominent research hotspots in the field of LLM. However, existing RAG frameworks for knowledge bases still face several challenges, including the inability to extract key information, incomplete answers, and inaccuracies, which may lead to fragmented answers and low-quality responses. To address these issues, this paper proposes a mixed RAG method that combines VectorRAG and GraphRAG to improve the accuracy and contextual relevance of knowledge base question answering. Additionally, to tackle the problem where user queries may lack specific entities, causing GraphRAG to fail to extract key information, we introduce a dual-retrieval paradigm. Experimental results validate the effectiveness of this approach in enhancing the model's question-answering abilities.