Search papers, labs, and topics across Lattice.
The paper introduces Emotion Collider (EC-Net), a hyperbolic hypergraph framework for multimodal emotion and sentiment modeling that uses Poincare-ball embeddings to represent modality hierarchies. EC-Net employs a hypergraph mechanism for bidirectional message passing between nodes and hyperedges, and contrastive learning in hyperbolic space with decoupled radial and angular objectives to improve class separation. Experiments on multimodal emotion benchmarks demonstrate that EC-Net generates robust representations and improves accuracy, especially under partial modality availability or noise.
Achieve robust multimodal sentiment analysis, even with noisy or incomplete data, by explicitly modeling hierarchical relationships in hyperbolic space.
Emotional expression underpins natural communication and effective human-computer interaction. We present Emotion Collider (EC-Net), a hyperbolic hypergraph framework for multimodal emotion and sentiment modeling. EC-Net represents modality hierarchies using Poincare-ball embeddings and performs fusion through a hypergraph mechanism that passes messages bidirectionally between nodes and hyperedges. To sharpen class separation, contrastive learning is formulated in hyperbolic space with decoupled radial and angular objectives. High-order semantic relations across time steps and modalities are preserved via adaptive hyperedge construction. Empirical results on standard multimodal emotion benchmarks show that EC-Net produces robust, semantically coherent representations and consistently improves accuracy, particularly when modalities are partially available or contaminated by noise. These findings indicate that explicit hierarchical geometry combined with hypergraph fusion is effective for resilient multimodal affect understanding.
