Integrating Retrieval-Augmented Generation (RAG) and Knowledge Augmented Generation (KAG) Frameworks to Build Accurate Enterprise Question Answering Systems

With the rapid development of Large Language Models (LLMs), more and more requirements from enterprises for accurate question answering (QA) systems are emerging. LLMs have shown great potential in natural language processing (NLP) but face limitations such as hallucinations and lack of domain-specific knowledge. Retrieval-Augmented Generation (RAG) addresses these issues by retrieving external information to enhance LLMs' outputs, while Knowledge Augmented Generation (KAG) further improves reasoning capabilities through knowledge graphs. We propose five strategies for combining RAG and KAG, each balancing storage, retrieval efficiency, and response accuracy. Experiments using a dataset from a power industry company show that integrating RAG and KAG improves QA system performance, with Strategy 5 achieving the highest accuracy rate of 78%. Our findings highlight the potential of combining retrieval and knowledge augmentation to enhance enterprise QA systems.