Abstract: Highlights•We propose a Similar Modality Completion based Multimodal Sentiment Analysis model for uncertain missing modalities.•We propose a set of strategies for completing the missing modalities for uncertain modalities missing scenarios.•We fuse the text’s representation into the video and audio with the guidance of a pre-trained model to enhance the quality of the video and audio.
Loading