基于多视角解耦融合的多模态会话情感识别

张婷婷,朱小飞

Multi-view decoupled fusion for multimodal emotion recognition in conversation

Tingting ZHANG,Xiaofei ZHU

表 5 框架核心模块在2个数据集上的消融实验结果

Tab.5 Ablation results of core modules in framework on two datasets

模型 IEMOCAP MELD Acc/% W-F1/% Acc/% W-F1/% w/o MCE 71.51 71.26 67.44 66.41 w/o S-GRU 72.53 72.74 67.89 66.73 w/o D-GRU 72.41 72.54 67.80 66.69 w/o S-&D-GRU 72.35 72.48 67.74 66.61 w/o MMDF 69.70 69.58 67.30 66.51 w/o TM 72.73 72.67 67.91 66.79 w/o HW 72.58 72.64 67.98 66.74 w/o TM & HW 72.27 72.49 67.85 66.66 MDF 72.87 73.05 68.06 66.92