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

Table 6

The top 5 and 10 recall on the test set.

Model	HFDS		TDS
	Top 5 recall	Top 10 recall	Top 5 recall	Top 10 recall
Encoder (Char)-Classification	0.9692 ± 0.0028	0.9818 ± 0.0013	0.8906 ± 0.0045	0.9164 ± 0.0037
Encoder (Word)-Classification	0.9635 ± 0.0025	0.9785 ± 0.0015	0.8928 ± 0.0046	0.9195 ± 0.0038
BERT-Classification	0.9758 ± 0.0012	0.9858 ± 0.0011	0.9212 ± 0.0019	0.9426 ± 0.0015

Note. The results are expressed as mean ± SD. , compared with BERT-Classification.