About Traditional Chinese model

[pactera-song]

Does spaCY support traditional Chinese? ZH-TW or ZH-HK.

[adrianeboyd]

In the basic Chinese language support, there's not much specific to simplified Chinese, except maybe the stop words. If you have a jieba dictionary or pkuseg model for traditional Chinese characters, it should work fine.

The provided Chinese models like zh_core_web_sm are trained on OntoNotes 5, which only contains simplified Chinese (see p. 27 in the docs: https://catalog.ldc.upenn.edu/docs/LDC2013T19/OntoNotes-Release-5.0.pdf). If you know of data with a permissive license that can be used to train models for traditional Chinese (typically it's hardest to find NER data), we'd be happy to look into whether we could provide additional models.