核技术知识图谱的构建研究

Research on Construction of Nuclear Technology Knowledge Graph

论文提出一种核技术知识图谱的构建方法,解决核技术知识不成体系、获取渠道难的问题,为用户提供一种智能化和结构化获取方式。首先采用自顶向下与自底向上相结合的方法分别构建模式层与数据层,利用Protégé本体编辑工具实现核技术知识图谱模式层的构建。使用爬虫技术实现核技术半结构化数据的抽取。根据ME+R+BIESO标注方法对核技术非结构化数据进行标注,分别使用CRF、BI-LSTM和BI-LSTM-CRF模型的命名识别方法对标注数据进行实体抽取,并通过实验结果得到BI-LSTM-CRF模型在核技术知识数据上效果优于其他两个模型。然后,通过知识融合,处理重复和错误的数据,实现核技术知识图谱数据层的构建。其次,采用Neo4j图数据库存储知识,成功构建核技术领域知识图谱。

This paper proposes a construction method of nuclear technology knowledge graph,which solves the problems of fragmentation of nuclear technology knowledge and difficult acquisition channels,and provides users with an intelligent and struc-tured acquisition method.Firstly,the top-down and bottom-up methods are used to construct the mode layer and data layer respec-tively,and the Protégéontology editing tool is used to construct the mode layer of nuclear technology knowledge graph.The crawler technology is used to extract semi-structured data.The unstructured data of nuclear technology is labeled according to the ME+R+BIESO labeling method.The naming and recognition methods of CRF,BI-LSTM and BI-LSTM-CRF models are used to extract en-tities from the labeled data respectively.Through the experimental results,it is obtained that the effect of BI-LSTM-CRF model on nuclear technology knowledge data is better than the other two models.Then,through knowledge fusion,duplicate and erroneous da-ta are processed to construct the data layer of the nuclear technology knowledge graph.Secondly,Neo4j graph database is used to store knowledge,and the knowledge graph of nuclear technology field is successfully constructed.