Phenonizer: A Fine-Grained Phenotypic Named Entity Recognizer for Chinese Clinical Texts

Table 8

The symptom extraction performance of models on isomorphic data (COVID-19).

Training dataset	COVID-19			TCM-HN
Models	Precision	Recall	F1-score	Precision	Recall	F1-score
BiLSTM-CRF	0.9128	0.9127	0.9128	0.7739	0.7675	0.7707
GloVeWiki-BiLSTM-CRF	0.9064	0.9116	0.9090	0.7626	0.7715	0.7670
GloVeMedical-BiLSTM-CRF	0.9093	0.9144	0.9113	0.7683	0.7661	0.7672
W2VWiki-BiLSTM-CRF	0.9145	0.9209	0.9177	0.7994	0.8380	0.8181
W2VMedical-BiLSTM-CRF	0.9164	0.9201	0.9183	0.8104	0.8457	0.8275
BERT-CRF	0.9220	0.9231	0.9225	0.8056	0.8440	0.8243
BERT-BiLSTM	0.9170	0.9220	0.9195	0.8188	0.8516	0.8348
Phenonizer	0.9211	0.9264	0.9237	0.8230	0.8556	0.8389