Natural Language Processing Algorithms for Normalizing Expressions of Synonymous Symptoms in Traditional Chinese Medicine
Table 4
Model performance on HFDS test data sets.
Model
Accuracy
Precision
Recall
F1-score
Encoder (Char)-Decoder (Char)
0.8641 ± 0.0065a,b
0.8656 ± 0.0084a,b
0.8555 ± 0.0062a,c
0.8605 ± 0.0056a,b
Encoder (Char)-Decoder (Label)
0.8558 ± 0.0070a,b
0.8727 ± 0.0038a,b
0.8463 ± 0.0062a,b
0.8593 ± 0.0043a,b
Encoder (Word)-Decoder (Label)
0.8487 ± 0.0046a,b
0.8678 ± 0.0076a,b
0.8377 ± 0.0054a,b
0.8525 ± 0.0059a,b
Encoder (Word)-Decoder (Word)
0.8451 ± 0.0035a,b
0.8472 ± 0.0056a,b
0.8345 ± 0.0036a,b
0.8408 ± 0.0023a,b
Encoder (Char)-Classification
0.8311 ± 0.0078c
0.8937 ± 0.0072c
0.8342 ± 0.0077c
0.8629 ± 0.0045c
Encoder (Word)-Classification
0.8294 ± 0.0079c
0.8983 ± 0.0055c
0.8302 ± 0.0070c
0.8629 ± 0.0038c
BERT-UniLM (Char)
0.8914 ± 0.0059c
0.8983 ± 0.0042c
0.8855 ± 0.0077c
0.8918 ± 0.0043c
BERT-UniLM (Label)
0.8829 ± 0.0046c
0.8909 ± 0.0069c
0.8773 ± 0.0044c
0.8840 ± 0.0036c
BERT-Classification
0.9051±0.0039
0.9118±0.0033
0.9028±0.0046
0.9073±0.0033
Note. The results are expressed as mean ± SD, and the threshold value of the sigmoid function was 0.2. a, compared with BERT-UniLM (Char); b: , compared with BERT-UniLM (Label). c, compared with BERT-Classification.