Temporal text-guided feedback-based progressive fusion network for multimodal sentiment analysis
The goal of text-guided multimodal sentiment analysis (MSA) is to fully leverage the unique sentiment advantages of textual features.However, existing research on multimodal sentiment analysis (MSA) predominantly relies on unidirectional text-dominant frameworks, neglecting the challenges posed by modality imbalance and sentiment conflicts in scena