Research on BERT-Based Audit Entity Extraction Method
Rui Xiang, Wei Bo Li, Hua Yan · 2021
Addresses the problems faced in extracting audit-specific knowledge and manual extraction of entities when using knowledge graph methods in intelligent auditing. A BERT-based entity extraction method is proposed. Firstly, word vectors with fused features are generated by word embedding using BERT, then the text information is extracted by BILSTM, and finally the final entities are extracted using CRF with the addition of sequence annotation links and rules. The experimental results show that the method achieves an F1 value of 98.03% on the audit meeting business dataset, which has good applicability and practicality.