Order embeddings and character-level convolutions for multimodal alignment

Re-assign community