Table 4.
Micro F1-scores for all classes and models with microaverage F1-scores and macro-average precision on the complete test data set.
| Type of trauma | Test data set (n=4315), n (%) | Micro F1-scores | ||||
|
|
|
TF-IDFa/SVMb | CamemBERT | FlauBERT | BelGPT2 | GPTanam |
| Accident of exposure to bodily fluids | 41 (1) | 0.83 | 0.84 | 0.84 | 0.83 | 0.91 c |
| Assault | 498 (11.54) | 0.9 | 0.91 | 0.92 | 0.91 | 0.96 |
| MVAd | 568 (13.16) | 0.91 | 0.90 | 0.91 | 0.91 | 0.97 |
| Foreign body in eye | 186 (4.3) | 0.79 | 0.84 | 0.82 | 0.82 | 0.97 |
| Fall | 1554 (36.01) | 0.9 | 0.92 | 0.91 | 0.92 | 0.98 |
| Sport accident | 371 (8.6) | 0.82 | 0.83 | 0.83 | 0.85 | 0.94 |
| Intentional injury | 112 (2.6) | 0.75 | 0.76 | 0.73 | 0.77 | 0.84 |
| Other trauma | 985 (22.8) | 0.8 | 0.83 | 0.82 | 0.85 | 0.98 |
| Micro F1-score | N/Ae | 0.864 | 0.878 | 0.873 | 0.887 | 0.969 |
| Macro precision | N/A | 0.860 | 0.880 | 0.880 | 0.89 | 0.970 |
aTF-IDF: term frequency–inverse document frequency.
bSVM: support vector machine.
cThe best F1-scores are in italic.
dMVA: motor vehicle accident.
eN/A: not applicable.