search

facebookresearch / MUSE

A library for Multilingual Unsupervised or Supervised word Embeddings
Other
3.17k stars 544 forks source link

Which cross-lingual vectors file I should use in NMT? #163

Open lihongzheng-nlp opened 4 years ago

lihongzheng-nlp commented 4 years ago

Hello @glample , I have a basic question about the trained cross-lingual vectors. Suppose I have obtained the (src)vectors-zh.txt and (tgt) vectors-vi.txt with Chinese and Vietnamese data, and I want to apply the vector file in zh2vi NMT as pretrained embeddings, so which exact file I should use? or both? As you mentioned in https://github.com/facebookresearch/MUSE/issues/121

The target embeddings are not moved, only the source are mapped to the target embeddings.

I guess I should use the src language vector file in zh2vi NMT, am I right? Thank you!