摘要
特定领域中专业术语的密集使用增加了理解和抽取事件要素的难度。为此,提出一种融合知识图谱与预训练模型的事件抽取方法KG-PMEE。在装备这一特定领域上,将知识图谱中的实体信息嵌入到模型训练过程中,以增强预训练模型对专业术语和上下文信息的理解能力,可提升模型在目标领域的适应性。实验表明,所提方法在装备领域数据集上的F1值相较于基线模型ERNIE提升2.97%,在公开数据集ACE2005上的性能也得到了提升,验证了引入知识图谱信息对提升领域事件抽取效果的有效性。
Intensive use of specialized terminology in specific fields increases the difficulty of understanding and extracting event elements.Therefore,a event extraction method KG-PMEE that integrates knowledge graph and pre trained model is proposed.In the specific field of equipment,embedding entity information from the knowledge graph into the model training process enhances the pre trained model's understanding of professional terminology and contextual information,and improves the model's adaptability in the target domain.The experiment shows that the proposed method has an F1 value improvement of 2.97%compared to the baseline model ERNIE on the equipment domain dataset,and its performance has also been improved on the publicly available dataset ACE2005,verifying the effectiveness of introducing knowledge graph information in improving domain event extraction performance.
作者
方睿
崔良中
FANG Rui;CUI Liangzhong(Naval University of Engineering,Wuhan 430033,China)
出处
《软件导刊》
2025年第9期62-69,共8页
Software Guide
基金
装备预先研究项目(30209040702)。
关键词
事件抽取
知识图谱
预训练模型
特定领域
深度学习
event extraction
knowledge graph
pre-trained model
specific domain
deep learning
