Research on Domain-Specific Knowledge Graph Based on the RoBERTa-wwm-ext Pretraining Model

Table 9

Comparative experimental results of named entity recognition.

Model	Precision (%)	Recall (%)	F1
RoBERTa-wwm-ext + BiLSTM + CRF	81.11	85.13	83.07
BiLSTM + CRF	81.51	84.63	83.04
BERT-wmm-ext + BiLSTM + CRF	80.96	84.90	82.89
BERT + BiLSTM + CRF	81.93	84.01	82.96
IDCNN	79.46	84.73	82.01