Findings of the Second Shared Task on Multimodal Machine Translation and Multilingual Image Description