Leveraging a Joint learning Model to Extract Mixture Symptom Mentions from Traditional Chinese Medicine Clinical Notes

Table 3

Joint learning between official BERT-base (Chinese) and fine-tuned BERT.

Model	Label embedding	F1-score	Precision	Recall
BERT-base (Chinese)	Without	0.7968	0.7998	0.7939
	With	0.8102	0.8119	0.8085
Fine-tuned BERT (ours)	Without	0.8016	0.8218	0.7823
	With	0.8216	0.8250	0.8183