Abstract: Highlights•We propose a Multimodal Sentiment Analysis model for uncertain missing modalities.•We translate the visual and audio to the text to improve modalities’ quality.•We apply a pre-trained model to supervise the model to handle uncertain missing data.
Loading