Quantum-inspired multimodal fusion for video sentiment analysis

We tackle the crucial challenge of fusing different modalities features for multimodal sentiment analysis. Mainly based on neural networks, existing approaches largely model interactions in an implicit and hard-to-understand manner. address this limitation with inspirations from quantum theory, which contains principled methods modeling complicated correlations. In our quantum-inspired framework, word interaction within a single modality across are formulated superposition entanglement respectively at stages. The complex-valued network implementation framework achieves comparable results to state-of-the-art systems two benchmarking video analysis datasets. meantime, we produce unimodal bimodal directly interpret entangled decision.