一种多模块图注意力机制和词库构建的电力敏感数据识别算法  

A Power Sensitive Data Recognition Algorithm With Multi-module Graph Attention Mechanism and Thesaurus Construction

作　　者：朱红勤 余璟 王黎明 许洪华 马楠 ZHU Hongqin;YU Jing;WANG Liming;XU Honghua;MA Nan(Nanjing Power Supply Company,State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210019,Jiangsu Province,China)

机构地区：[1]国网江苏省电力有限公司南京供电分公司,江苏省南京市210019

出　　处：《电力信息与通信技术》2024年第7期27-34,共8页Electric Power Information and Communication Technology

基　　金：国网江苏省电力有限公司科技项目资助“电力调控数据安全监测及风险预警关键技术研究”(J2023109)。

摘　　要：在大数据时代下,电力行业的数据共享和融合存在巨大的风险。为此,文章提出了一种基于多模块图注意力机制的新文本分类算法。首先将图注意力机制与Transformer相结合,引入句法关系并运用图注意力机制提取文本特征;其次,为了提高敏感词识别准确性以及改善分词工具对敏感词的识别,提出了一种新型敏感词库构建方法。实验证明,文章提出的方法在敏感文本分类任务中表现出色,同时具备良好的可扩展性。In the era of big data,there is a great risk of data sharing and fusion in the power industry.For this reason,this paper proposes a new text categorization algorithm based on multi-module graph attention mechanism.Firstly,the graph attention mechanism is combined with Transformer to introduce syntactic relations and use the graph attention mechanism to extract text features.Secondly,in order to increase the accuracy of sensitive word recognition and improve the recognition of sensitive words by the word separation tool,this paper proposes a new method for constructing sensitive word base.Experiments prove that the proposed method performs well in the sensitive text classification task and has good scalability at the same time.

关 键 词：敏感数据 词库构建和扩展 智能识别和分类 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]