Table 5 Experimental results of the Dutch dataset.

From: Enhanced effective convolutional attention network with squeeze-and-excitation inception module for multi-label clinical document classification

Model

Dutch

# Labels

Precision

Recall

F1

XLM-RoBERTa

50

0.725

0.289

0.413

MultiResCNN

50

0.458

0.639

0.534

EffectiveCAN

50

0.822

0.760

0.790

EffectiveCAN (3500)

50

0.873

0.777

0.822

EffectiveCAN (3500)

144

0.844

0.732

0.784

EECAN (Multi-layer attention)

55

0.882

0.786

0.836

EECAN (Sum-pooling attention)

150

0.852

0.745

0.795

Back to article page