Hierarchical Contrastive Attention and Dynamically Heterogeneous Fusion-Based Multimodal Sentiment Synergistic Analysis Model

Multimodal sentiment analysis plays a crucial role in intelligent human-computer interaction, personalized recommendation systems, and psychological health monitoring. However, existing models still struggle with modality heterogeneity, dynamic interaction modeling, and sentiment feature extraction. To address these limitations, this paper proposes a Hierarchical Contrastive Attention and Dynamically Heterogeneous Fusion-Based Multimodal Sentiment Synergistic Analysis Model, which employs a dual-network framework to enhance cross-modal information fusion. The experimental results demonstrate that HCD-MSA achieves significantly higher accuracy, recall, and F1-score compared to existing models on both the CMU-MOSEI and IEMOCAP datasets. Furthermore, ablation studies confirm the effectiveness of each module. This research offers new insights into multimodal sentiment analysis, providing both theoretical support and practical guidance for cross-modal information fusion and sentiment recognition.