search

salesforce / cove

BSD 3-Clause "New" or "Revised" License
473 stars 93 forks source link

how to train to get Chinese embeddings. #17

Closed skyhiter closed 5 years ago

skyhiter commented 6 years ago

hi, i want to use CoVe model to Chinese corpus, so i need to train MT-LSTM using Chinese to English Parallel Corpus. how can i do that?

thanks.

bmccann commented 5 years ago

You can use any open source MT system with Chinese to English parallel data, and the encoder's outputs are a version of CoVe. I used OpenNMT-py for the pretrained weights in this repo.

skyhiter commented 5 years ago

copy that. Thanks.