基于结构化知识图谱检索增强的电力领域长文本生成OA

In response to the problems of lack of professional knowledge,illusion generation,and insufficient contextual consistency in the long-text generation in in the power field,this paper proposes a method for generating long-text based on structured knowledge graph retrieval-augmented.Firstly,through graph reinforcement learning,key entities and relationships are searched in the knowledge graph,and accurate knowledge retrieval is achieved by integrating local and global features;Then,a global local multi-agent large model framework is constructed.The global model determines the overall structure by generating summaries and directories,while the local model generates chapter content based on retrieved structured knowledge.Through multi-agent iterative interaction,the professionalism and coherence of long-text are optimized.The experimental results show that the proposed method significantly improves accuracy and consistency in long-text generation tasks in the power field,with a 4.5%improvement in generation quality compared to general large models.It outperforms the comparison method in indicators such as ROUGE-2 and Jaccard similarity,providing an effective solution for intelligent processing in scenarios such as power system operation and maintenance,fault diagnosis,and power office,and helping the digital transformation of the power industry.