232 - Improving Video Captioning with Temporal Composition of a Visual-Syntactic Embedding