Hi Is there any analysis on how this model performs on non-Latin langauges such as Chinese. Japanese, Korean etc?