The obvious answer would appear to be that the algorithm was likely trained on large tranches of Chinese data, and, thus, ...