Table 6 Impact of the different character embeddings in our method
From: Multi-task learning for Chinese clinical named entity recognition with external knowledge
Dataset | Method | Precision | Recall | F-measure |
|---|---|---|---|---|
CCKS2017 | Random | 89.56 | 89.29 | 89.42 |
BERT(Baseline1) | 90.73 | 90.51 | 90.62 | |
FT-BERT(Baseline2) | 91.27 | 91.21 | 91.24 | |
FT-BERT-Radical | 91.69 | 91.34 | 91.51 | |
FCCd | Random | 84.23 | 83.32 | 83.77 |
BERT(Baseline1) | 86.11 | 85.52 | 85.81 | |
FT-BERT(Baseline2) | 86.21 | 85.83 | 86.02 | |
FT-BERT-Radical | 86.95 | 86.56 | 86.75 |