Table 6 Ablation experiment on the feature fusion module at the sentence level in ZuCo.
From: Multi-branch convolutional neural network with cross-attention mechanism for emotion recognition
Feature fusion | Acc (%) | Pre (%) | Rec (%) | F1 (%) |
|---|---|---|---|---|
TDF+Text | 65.69 (-29.95) | 61.73 (-33.96) | 49.18 (-46.36) | 57.74 (-37.86) |
FDF+Text | 63.55 (-28.56) | 65.03 (-27.06) | 55.99 (-36.05) | 55.99 (-35.94) |
TFDF+Text | 91.89 (-04.29) | 92.09 (-04.17) | 91.78 (-04.25) | 91.93 (-04.18) |
TDF+FDF+Text | 90.25 (-05.78) | 90.44 (-05.34) | 90.27 (-05.27) | 90.35 (-05.69) |
TDF+TFDF+Text | 95.08 (-01.24) | 95.22 (-00.93) | 94.94 (-01.28) | 94.94 (-01.28) |
FDF+TFDF+Text | 93.24 (-03.00) | 93.32 (-02.94) | 93.23 (-02.00) | 93.27 (-02.93) |
TDF+FDF+TFDF+Text | 95.32 (-01.63) | 95.15 (-01.83) | 95.22 (-01.83) | 95.18 (-01.83) |
