stt model training question

NVIDIA / OpenSeq2Seq

Toolkit for efficient experimentation with Speech Recognition, Text2Speech and NLP

https://nvidia.github.io/OpenSeq2Seq

Apache License 2.0

1.54k stars 369 forks source link

Closed wujsy closed 5 years ago

wujsy commented 5 years ago

Hi, for speech to text in chinese ,what are the differences between alphabet.txt and vocabulary.txt? Are they both necessary?

wujsy commented 5 years ago

How to generate the trie file for chinese ?

borisgin commented 5 years ago

See for example https://blog.yuwu.me/?p=3989 , starting from 11.

GabrielLin commented 5 years ago

@borisgin Would you please share your CER or WER on THCHS-30? Thanks.

borisgin commented 5 years ago

We did not train ASR for Chinese

mcggood commented 4 years ago

See for example https://blog.yuwu.me/?p=3989 , starting from 11.

the link can't be reached. Would you share again?