面向微波组件故障文本的知识抽取方法

本研究的主要目标是在微波技术领域应用知识抽取技术优化微波组件的调试和装配过程。鉴于知识抽取技术在微波技术领域的不足，文中提出了一种基于CasRel-LN模型的知识抽取算法。该模型综合运用了BERT编码器、主体标记模型和特定关系下客体标记模型，成功实现了对微波组件故障文本的实体抽取。实验结果表明，CasRel-LN模型在测试集上的综合性能优于其他知识抽取模型，提高了实体抽取的准确率和召回率。使用该算法构建的微波组件故障知识图谱包含1 568个实体和1 618条三元组，存储在Neo4j图数据库中。通过知识图谱的可视化展示，调试人员可以更高效地解决微波组件调试中的复杂问题，确保电子设备的稳定运行，最终为微波组件在装配阶段就能达到更好的性能提供更有效的支持。

Knowledge Extraction Method for Microwave Component Fault Text

The primary objective of this study is to apply knowledge extraction techniques in the field of microwave technology to optimize the debugging and assembly processes of microwave components. For inadequacies of knowledge extraction techniques in the domain of microwave technology, a knowledge extraction algorithm based on the CasRel-LN model is proposed. This model comprehensively incorporates the BERT encoder, subject tagging model and object tagging model under specific relationship contexts and successfully achieves entity extraction of microwave component fault texts. Experimental results demonstrate that the CasRel-LN model, which enhances the precision and recall of entity extraction, exhibits superior overall performance on the test set compared to other knowledge extraction models. A knowledge graph of microwave component faults is constructed by this algorithm. This knowledge graph comprises 1 568 entities and 1 618 triplets, which are stored in the Neo4j graph database. Through the visual representation of the knowledge graph, debugging personnel can efficiently solve complex issues in the debugging of microwave components and thus ensure the stable operation of electronic devices. Fault extraction and debugging can provide more effective support for achieving better performance during the assembly phase.