Table 4 Performance at entity-level and token-level
From: Leveraging text skeleton for de-identification of electronic medical records
Model | Entity-level | Token-level | ||||
|---|---|---|---|---|---|---|
Precision | Recall | F1-score | Precision | Recall | F1-score | |
Rule-based | 0.8747 | 0.9276 | 0.9003 | 0.8802 | 0.9478 | 0.9128 |
CRF | 0.9815 | 0.8972 | 0.9375 | 0.9669 | 0.9236 | 0.9448 |
Bi-LSTM | 0.9701 | 0.9235 | 0.9462 | 0.9545 | 0.9027 | 0.9279 |
Bi-GRU | 0.9665 | 0.9470 | 0.9567 | 0.9592 | 0.9270 | 0.9428 |
TS-GRU | 0.9778 | 0.9502 | 0.9638 | 0.9777 | 0.9447 | 0.9609 |