The zh dataset seems not right.
#2
by
dedongli
- opened
The zh dataset is not right I think. In my opinion, it should be pinyin, right?
Pinyin, not phonemes?
Some languages such as Chinese, Japanese and Korean require a different approach; espeak-ng isn't good with every language.
Hmm, ok. Do you have any suggestions? I think STTS might want phonemes, is there any good Chinese -> Phoneme library?
Sorry for the late reply.
I'm not familiar with the Chinese Pre-processing steps and the common libraries they use. there should be someone in the Slack server that can help with this.