Table 6 Training epochs per model (average of five rounds and standard deviation)
From: Hybrid natural language processing tool for semantic annotation of medical texts in Spanish
Model | UMLS | Medication | Temporal | Negation/ | Miscellaneous | Experiencer/ |
|---|---|---|---|---|---|---|
| Â | entities | data | entities | Speculation | entities | Temporality |
Bi-LSTM-CRF | 95.60 | 68.40 | 68.20 | 80.60 | 78.80 | 71.20 |
|  | (± 5.37) | (± 7.47) | (± 7.29) | (± 15.07) | (± 9.44) | (± 7.73) |
RoBERTa EHR | 17.00 | 14.20 | 14.00 | 10.80 | 16.80 | 10.80 |
|  | (± 2.83) | (± 3.63) | (± 2.24) | (± 1.92) | (± 3.56) | (± 4.09) |
EriBERTa | 15.50 | 10.40 | 12.80 | 17.25 | 15.60 | 16.40 |
|  | (± 4.12) | (± 3.78) | (± 6.72) | (± 5.50) | (± 6.07) | (± 3.36) |
CLIN-X-ES | 17.00 | 14.60 | 13.60 | 21.80 | 19.40 | 16.40 |
|  | (± 2.83) | (± 3.65) | (± 4.22) | (± 5.36) | (± 6.07) | (± 4.83) |
mBERT | 14.75 | 11.40 | 14.00 | 16.00 | 18.20 | 11.60 |
|  | (± 6.18) | (± 5.22) | (± 5.61) | (± 6.16) | (± 2.49) | (± 2.07) |
mDeBERTa vs 3 | 17.80 | 11.60 | 9.80 | 15.00 | 18.00 | 11.60 |
|  | (± 4.92) | (± 5.68) | (± 2.28) | (± 6.93) | (± 2.74) | (± 4.83) |