Optimization of Knowledge Graph-Driven Text Generation 
Methods within the RAG Framework

Qiaoyang Zheng

doi:10.70711/aitr.v3i3.8039

Optimization of Knowledge Graph-Driven Text Generation Methods within the RAG Framework

Qiaoyang Zheng

Abstract

Retrieval-Augmented Generation (RAG) frameworks demonstrate significant advantages in knowledge-intensive tasks, but traditional approaches fall short in handling complex knowledge associations and ensuring factual consistency. This paper proposes an optimized
RAG method integrating knowledge graphs by establishing a multimodal knowledge retrieval mechanism and a graph-structured information encoding strategy, achieving deep fusion between structured knowledge and generative models. The proposed method designs an entityrelation encoder based on graph neural networks, constructs a text-graph joint retrieval architecture, and introduces a knowledge-constrained
decoding mechanism to ensure the accuracy of generated content. Experimental results show that compared to traditional RAG methods, this
approach improves factual accuracy by 15.2% and semantic coherence by 12.8%, while maintaining high generation efficiency.

Keywords

Retrieval-Augmented Generation; Knowledge Graph; Graph Neural Network; Text Generation

Full Text:

PDF

Included Database

References

[1] Yi Longwangjie, Yang Xiaotong, Zhang Saisai Reshaping the Intelligent Publishing Industry Based on Retrieval-Augmented Generation

(RAG) [J]. China Media Technology, 2025 (7): 19-24.

[2] Zhang Qiang, Liu Feng MDKG: RAG Framework Based on Multimodal Knowledge Graph [J]. Computer Application Abstracts, 2025,

41 (2): 182-184.

[3] Duan Jianyong, Lu Chaoyang, Wang Hao, etc A Semantic knowledge driven keyword extraction method for paper abstracts [J]. Intelligence Engineering, 2022, 8 (3): 3-12.

[4] Lewis, P., Perez, E., Piktus, A., et al. Dense Passage Retrieval for Open-Domain Question Answering [EB/OL]. arXiv preprint

arXiv:2004.04906, 2020.

[5] Wang, Cunxiang, Haofei Yu, and Yue Zhang. "RFiD: Towards rational fusion-in-decoder for open-domain question answering." arXiv

preprint arXiv:2305.17041 (2023).

[6] Gao, Y., Xiong, Y., Gao, X., et al. Retrieval-Augmented Generation for Large Language Models: A Survey [EB/OL]. arXiv preprint

arXiv:2312.10997, 2024.

[7] Izacard, G., Lewis, P., Grave, E., et al. Leveraging Passage Retrieval with Generative Models for Open-Domain Question Answering [J].

Transactions of the Association for Computational Linguistics, 2021, 9: 1084-1096.

[8] Ji, S., Pan, S., Cambria, E., Marttinen, P., & Yu, P. S. A Survey on Knowledge Graphs: Representation, Acquisition and Applications [J].

IEEE Transactions on Knowledge and Data Engineering, 2021, 33(12): 1-1.

DOI: http://dx.doi.org/10.70711/aitr.v3i3.8039

Refbacks

There are currently no refbacks.