A Hybrid Model Based on Deep Convolutional Network for Medical Named Entity Recognition

Table 5

The performance of all models tested on the constructed dataset.

Model	Precision (%)	Recall (%)	F1-score (%)	FLOPs	Time (ms)
BiLSTM-CRF	96.00	96.10	96.05	65.45G	3.27
BERT	96.32	95.81	96.06	528.93G	52.73
BERT-CRF	96.75	96.12	96.43	541.21G	53.26
BERT-BiLSTM-CRF	96.60	96.33	96.47	561.62G	54.89