Natural Language Processing Algorithms for Normalizing Expressions of Synonymous Symptoms in Traditional Chinese Medicine
Table 5
Model performance on TDS test data sets.
Model
Accuracy
Precision
Recall
F1-score
Encoder (Char)-Decoder (Char)
0.7980 ± 0.0078a,b
0.7876 ± 0.0147a,b
0.7678 ± 0.0088a,b
0.7775 ± 0.0106a,b
Encoder (Char)-Decoder (Label)
0.7974 ± 0.0050a,b
0.8060 ± 0.0081a,b
0.7690 ± 0.0062a,b
0.7870 ± 0.0056a,b
Encoder (Word)-Decoder (Label)
0.7892 ± 0.0069a,b
0.7979 ± 0.0100a,b
0.7595 ± 0.0066a,b
0.7782 ± 0.0074a,b
Encoder (Word)-Decoder (Word)
0.7904 ± 0.0079a,b
0.7805 ± 0.0122a,b
0.7594 ± 0.0078a,b
0.7698 ± 0.0092a,b
Encoder (Char)-Classification
0.7559 ± 0.0056c
0.8560 ± 0.0125c
0.7278 ± 0.0057c
0.7866 ± 0.0058c
Encoder (Word)-Classification
0.7652 ± 0.0042c
0.8557 ± 0.0065c
0.7364 ± 0.0038c
0.7915 ± 0.0028c
BERT-UniLM (Char)
0.8274 ± 0.0087c
0.8152 ± 0.0115c
0.8043 ± 0.0082c
0.8097 ± 0.0094c
BERT-UniLM (Label)
0.8248 ± 0.0045c
0.8230 ± 0.0066c
0.7970 ± 0.0056c
0.8098 ± 0.0037c
BERT-Classification
0.8568±0.0029
0.8870±0.0039
0.8298±0.0037
0.8574±0.0026
Note. The results are expressed as mean ± SD, and the threshold value of the sigmoid function was 0.1. a, compared with BERT-UniLM (Char); b, compared with BERT-UniLM (Label). c, compared with BERT-Classification.