Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

Key Points

• The aim is to analyze multimodal language interactions using the CMU-MOSEI dataset with dynamic fusion graphs.
• Utilized the CMU-MOSEI dataset for multimodal language analysis.
• Employed dynamic fusion graphs for interpretable model development.
• Conducted evaluations to assess the effectiveness of the proposed methods.
• Dynamic fusion graphs significantly improved language understanding performance.
• Interpretation of multimodal interactions was enhanced through the proposed model.
• Results indicate practical applicability in real-world language processing tasks.