论文解读VISAFF: Speaker-Centered Visual Affective Feature Learning for Emotion Recognition in Conversation对话情感识别 | 7.4/10