Zhilu Yang

Factorize, Reconstruct, Enhance: A Unified Framework for Multimodal Sentiment Analysis

Multimodal Sentiment Analysis (MSA) aims to comprehensively and robustly interpret human emotions by integrating information from verbal, visual and acoustic modalities. However, the performance of existing models is often hampered by two key challenges: insufficient multilayer semantic extraction i