Research on feature-enhanced model for entity recognition of safety accident text based on deep learning
In order to study the influence of context semantic reference and complex domain content in the safety accident case reports on the performance of machine automatic recognition andinformation extraction,a BERT+Multi-CNN+BiGRU+CRF(BMulCBC)model considering the local feature enhancement was constructed.BERT was responsible for transforming the unstructured text into input,Multi-CNN and BiGRUwere responsible for encoding the vector local features and sequence features,and CRF was responsible for accuratelydecoding the entity labels.The results show that the precision rate,recall rate and Fl value of entity recognition of the model are 65.94％,74.02％and 69.75％,respectively,andthe precision rate and Fl value of the model are better than those of the similar comparison models.The research results can provide theoretical sup-port for the reasoning of the downstream safety accident event graph.
NETL
NSTL
万方数据
