Abstract: Modeling intra-modal and cross-modal interactions poses significant challenges in multimodal sentiment analysis. Currently, graph-based methods like HGraph-CL achieve promising performance, ...

The GCN models were trained in an unsupervised way, using the Deep Graph Infomax algorithm, and the proposed approaches employ recent rank-based manifold learning methods. Multimodal information is ...