What would be the best approach to align multiple (more than 2) monolingual word embeddings into a single vector space? I come from fastText_multilingual which is unfortunately outdated. With this repo you could get 70+ words embeddings aligned into a single vector space.
Hi,
What would be the best approach to align multiple (more than 2) monolingual word embeddings into a single vector space? I come from fastText_multilingual which is unfortunately outdated. With this repo you could get 70+ words embeddings aligned into a single vector space.
I also read this article from Facebook talking about how they merged lot of different word embeddings: https://engineering.fb.com/ml-applications/under-the-hood-multilingual-embeddings/
I would like to be able to do the same, not necessary for that much of languages, but at least a dozen. Any tip?
Thanks!