基于多头自注意力机制与MLP-Interactor的多模态情感分析

林宜山,左景,卢树华

Multimodal sentiment analysis based on multi-head self-attention mechanism and MLP-Interactor

Yishan LIN,Jing ZUO,Shuhua LU

表 2 在CMU-MOSI数据集上和其他基准模型性能的对比结果

Tab.2 Comparison of performance on CMU-MOSI dataset with other benchmark models

模型 MAE corr A2 A7 F1 TFN[40](2017) 0.901 0.698 — / 80.8 34.9 —/ 80.7 LMF[41](2018) 0.917 0.695 — / 82.5 33.2 —/ 82.4 MFN[42](2018) 0.965 0.632 77.4 / — 34.1 77.3 / — MulT[20](2019) 0.871 0.698 — / 83.0 40.0 — / 82.8 BBFN[11](2021) 0.776 0.755 —/84.3 45.0 —/84.3 Self-MM[24](2021) 0.713 0.798 84.0/85.98 — 84.42/85.95 MISA[7](2020) 0.783 0.761 81.8/83.4 42.3 81.7/83.6 MAG-BERT[43](2020) 0.731 0.798 82.5/84.3 — 82.6/84.3 CubeMLP[31](2022) 0.770 0.767 —/ 85.6 45.5 —/85.5 PS-Mixer[30](2023) 0.794 0.748 80.3/82.1 44.31 80.3/82.1 MTSA[44](2022) 0.696 0.806 —/86.8 46.4 —/86.8 AOBERT[10](2023) 0.856 0.700 85.2/85.6 40.2 85.4/86.4 TETFN[25](2023) TMRN[45](2023) 0.717 0.704 0.800 0.784 84.05/86.10 83.67/85.67 — 48.68 83.83/86.07 83.45/85.52 MTAMW[46](2024) 0.712 0.794 84.40/86.59 46.84 84.20/86.46 MIBSA[47](2024) 0.728 0.798 —/87.00 43.10 —/87.20 FRDIN[48](2024) 0.682 0.813 85.8/87.4 46.59 85.3/87.5 CRNet[49](2024) 0.712 0.797 —/86.4 47.40 —/86.4 本文模型 0.575 0.868 87.6/89.6 52.23 87.7/89.6