Table 4 Performance comparison between our method and baselines on MOSEI Dataset.
From: Multimodal sentiment analysis based on multi-layer feature fusion and multi-task learning
Models | MOSEI | Â | |||
|---|---|---|---|---|---|
Acc2 | F1-score | MAE | Corr | Acc7 | |
EF-\(\hbox {LSTM}^\textrm{a}\) | 80.1/80.3 | 80.3/81.0 | 0.603 | 0.682 | 49.3 |
LF-\(\hbox {DNN}^\textrm{a}\) | 78.6/82.3 | 79.0/82.2 | 0.561 | 0.723 | 52.1 |
\(\hbox {TFN}^\textrm{b}\) | –/82.5 | –/82.1 | 0.593 | 0.700 | 50.2 |
\(\hbox {LMF}^\textrm{b}\) | –/82.0 | –/82.1 | 0.623 | 0.677 | 48.0 |
\(\hbox {MFN}^\textrm{b}\) | 76.0/– | 76.0/– | – | – | – |
Graph-\(\hbox {MFN}^\textrm{a}\) | 81.9/84.0 | 82.1/83.8 | 0.569 | 0.725 | 51.9 |
\(\hbox {MulT}^\textrm{b}\) | –/82.5 | -/82.3 | 0.58 | 0.703 | 51.8 |
\(\hbox {MISA}^\textrm{b}\) | 83.6/85.5 | 83.8/85.3 | 0.555 | 0.756 | 52.2 |
Self-\(\hbox {MM}^*\) | 77.6/83.9 | 78.5/84.0 | 0.540 | 0.757 | 54.2 |
\(\hbox {MMIM}^*\) | 82.1/83.8 | 82.3/83.6 | 0.599 | 0.740 | 51.9 |
Our Method | 84.8/86.0 | 84.9/85.8 | 0.535 | 0.760 | 54.5 |
