Machine Translation

In week 3, there is a lot of discussion about translating from French to English. My doubt is, just like there were word embedding learnt for English vocabulary which was used in LSTM, etc., are there word embedding learnt for French vocabulary as well to be used in these architectures?
If yes, then is it the same for all the other languages?
If no, then how does it work well without word embedding in the other language?

thanks for answering

You could have an embedding for each language.

Or you could have a multi-language embedding.

https://www.cs.cmu.edu/~afm/projects/multilingual_embeddings.html