Multi-level textual-visual alignment and fusion network for multimodal aspect-based sentiment analysis | Synapse