As we noted on Saturday, with the release of the Visual Knowledge Graph (VGKG) 1.0, we are now passing language hints to the Google Cloud Vision API estimating the likely language(s) it should expect to see of any text that might be present in each image. This has had a dramatic impact on the recognition of Arabic, Persian, Thai, Chinese and other script and non-Latin languages. We're tremendously excited about the potential of this massive new range of languages we are able to OCR.