Noisy Image Captions

We present a dataset and method for improving the translation of noisy image captions that were created by users of Wikimedia Commons. The dataset is multilingual but non-parallel, and is …...