Natural Language Processing Algorithms for Normalizing Expressions of Synonymous Symptoms in Traditional Chinese Medicine

Table 3

Model parameters and development set results on TDS.

Model	LR	DR	MC	Accuracy	Precision	Recall	F1-score
Encoder (Char)-Decoder (Char)	0.0005	0.3	512	0.8212 ± 0.0038	0.8307 ± 0.0107	0.8011 ± 0.0044	0.8156 ± 0.0067
Encoder (Char)-Decoder (Label)	0.0003	0.5	512	0.8160 ± 0.0026	0.8379 ± 0.0060	0.7959 ± 0.0030	0.8164 ± 0.0033
Encoder (Word)-Decoder (Label)	0.0003	0.5	512	0.8091 ± 0.0045	0.8320 ± 0.0048	0.7875 ± 0.0052	0.8091 ± 0.0040
Encoder (Word)-Decoder (Word)	0.0001	0.3	256	0.8053 ± 0.0028	0.8167 ± 0.0076	0.7838 ± 0.0034	0.7999 ± 0.0049
Encoder (Char)-Classification	0.01	0.5	512	0.7681 ± 0.0058	0.8876 ± 0.0114	0.7503 ± 0.0051	0.8132 ± 0.0061
Encoder (Word)-Classification	0.01	0.3	512	0.7790 ± 0.0051	0.8913 ± 0.0074	0.7595 ± 0.0060	0.8201 ± 0.0034
BERT-UniLM (Char)	0.00003	0.1	N/A	0.8338 ± 0.0027	0.8399 ± 0.0047	0.8180 ± 0.0034	0.8288 ± 0.0033
BERT-UniLM (Label)	0.00003	0.1	N/A	0.8219 ± 0.0017	0.8388 ± 0.0056	0.8018 ± 0.0034	0.8199 ± 0.0028
BERT-Classification	0.00003	0.1	N/A	0.8547 ± 0.0027	0.9072 ± 0.0037	0.8405 ± 0.0026	0.8726 ± 0.0024

Note. LR: learning rate; DR: dropout rate; MC: number of memory cells of RNN; N/A: not applicable.