Self-adaptive attention fusion for multimodal aspect-based sentiment analysis
Multimodal aspect term extraction (MATE) and multimodal aspect-oriented sentiment classification (MASC) are two crucial subtasks in multimodal sentiment analysis.The use of pretrained generative models has attracted increasing attention in aspect-based sentiment analysis (ABSA).However, the inherent semantic gap between textual and visual modalitie