企业审计知识图谱的构建——以华神科技公司为例
The Construction of the Enterprise Audit Knowledge Graph—Taking Huasen Technology Company as an Example

应对传统审计无法满足日益增长的全方位精确审计需求的挑战，智能审计相关技术应运而生，但仍处于弱智能化阶段，未能达到可用于审计知识推理的标准。本文基于BERT-BiLSTM-CRF实体识别模型，创新性地运用了改进的预训练模型BERT-WWM，提出了新的企业内控审计知识图谱的构建方法，获得最优序列标注，对审计报告、财务报表及内部审计文件等实体进行字符抽取，在知识图谱中引入深度学习、有效挖掘海量审计实体之间的复杂关系，并实现不同结构数据的融合。本文以华神科技公司为例，采用neo4j图数据库构建审计知识图谱，挖掘知识图谱推理价值、延拓审计知识图谱使用的广度和深度，为审计知识推理赋能，并为企业创造更高价值。
To address the challenges of increasingly comprehensive and precise auditing demands that traditional auditing methods cannot meet, intelligent auditing technologies have emerged. However, these technologies are still in a stage of weak intelligence and have not yet reached the standards required for knowledge reasoning in auditing. In this paper, based on the BERT-BiLSTM-CRF entity recognition model, we innovatively utilize an improved pre-training model called BERT-WWM and propose a new method for constructing an enterprise internal control audit knowledge graph. This method achieves optimal sequence labeling by extracting characters from entities such as audit reports, financial statements, and internal audit docu-ments. We introduce deep learning into the knowledge graph and effectively mine complex relationships among massive auditing entities, realizing the fusion of different structured data. Tak-ing Huashen Technology Company as an example, we use the neo4j graph database to construct an audit knowledge graph, explore the inference value of the knowledge graph, expand the breadth and depth of its use in audit knowledge reasoning, empower audit knowledge reason-ing, and create higher value for enterprises.