电力信息与通信技术2025，Vol.23Issue(5)：61-67,7.DOI:10.16543/j.2095-641x.electric.power.ict.2025.05.08
基于大模型的大规模电力数据零样本实体关系抽取方法
Zero-shot Entity Relationship Extraction From Massive Electric Power Data With Large Language Model
摘要
Abstract
In the era of big data,the extraction of entity relations triples from data in the electric power field is of great signifi-cance to the structuring of electric power data.However,traditional deep learning methods require a large amount of labeled training data in the power field for deep learning model learning.This method is undoubtedly time-consuming and labor-intensive,and cannot achieve good entity relations extraction results.In response to these problems,this pa-per proposes a zero-shot entity relations extraction method in the electric power field called chat relation exaction(ChatRE)based on a large language model.The two-stage multi-round dialogue method proposed by ChatRE divides the entity relations extrac-tion into two sub-tasks and uses syntactic enhancement.This further improves the extraction effect and effectively solves the above problems.This paper conducts an experimental evaluation of ChatRE based on the constructed elec-tric power field entity data set and general data set.The results show that the ChatRE model proposed in this paper is better than existing methods and can effectively extract entity relations triples from complex power domain knowledge.关键词
电力数据/大语言模型/零样本学习/实体关系抽取/多轮对话/句法分析Key words
electric power data/large language model/zero-shot learning/entity relations extraction/multi-turn dialogue/syntactic analysis分类
信息技术与安全科学引用本文复制引用
胡新雨,宋博川,仝杰,李云鹏,毛艳芳,吕晓祥,张强,孙大军,陈群丰..基于大模型的大规模电力数据零样本实体关系抽取方法[J].电力信息与通信技术,2025,23(5):61-67,7.基金项目
国家电网有限公司总部科技项目"电网主设备运维多模态生成式模型构建与应用关键技术研究"(5700-202318598A-3-2-ZN). （5700-202318598A-3-2-ZN）
