摘要
工程保障实体关系抽取对于开展智能化辅助决策具有重要意义。本文在领域专家的指导下设计了工程保障实体关系,并本地化部署DeepSeek,经过TCL、CoT多轮Prompt学习,抽取了工程保障实体关系;对该领域的数据集进行人工标注,开展基于四种模型的关系抽取实验,结果表明,RoBERTa-BiGRU-ATT模型的实体关系抽取效果最好,精确率、召回率和F1值分别达到了0.8903、0.8879、0.8885。本文研究表明,基于DeepSeek和DeepLearning融合策略对抽取实体关系,是一种更高效便捷的方法,对于构建领域知识图谱和智能问答具有重要意义。
Engineering support entity relation extraction is of great significance for intelligent assistant decision-making. In this paper, under the guidance of domain experts, the engineering support entity relationship is designed, and DeepSeek is deployed locally. After TCL and CoT rounds of Prompt learning, the engineering support entity relationship is extracted. The data sets in this field are manually labeled, and relationship extraction experiments based on four models are carried out. The results show that the RoBERTa-BiGRU-ATT model has the best entity relationship extraction effect, and the accuracy rate, recall rate and F1 value are 0.8903,0.8879 and 0.8885, respectively. The research in this paper shows that the fusion strategy based on DeepSeek and DeepLearning is a more efficient and convenient method for extracting entity relationships, which is of great significance for constructing domain knowledge graphs and intelligent question answering.
出处
《数据挖掘》
2025年第3期242-253,共12页
Hans Journal of Data Mining
