Robust-MSA: Understanding the Impact of Modality Noise on Multimodal Sentiment Analysis

• DOI: 10.1609/aaai.v37i13.27078
• Publication Date: 2023-06-27T18:46:58Z
• Authors: Huisheng Mao; Baozheng Zhang; Hua Xu; Ziqi Yuan; Yihe Liu
• Keywords: Robustness; Modality (human–computer interaction); Multimodal therapy

Improving model robustness against potential modality noise, as an essential step for adapting multimodal models to real-world applications, has received increasing attention among researchers. For Multimodal Sentiment Analysis (MSA), there is also a debate on whether are more effective noisy features than unimodal ones. Stressing intuitive illustration and in-depth analysis of these concerns, we present Robust-MSA, interactive platform that visualizes the impact noise well simple defence methods help researchers know better about how their perform with imperfect data.