Pre-trained subword embeddings in 275 languages, based on Byte-Pair Encoding (BPE) - View it on GitHub