Video-level sentiment analysis is a challenging task and requires systems to obtain discriminative multimodal representations that can capture difference in sentiments across various modalities. However. due to diverse distributions of various modalities and the unified multimodal labels are not always adaptable to unimodal learning. the distance difference between unimodal representa... https://www.animationbengal.com/