Pretrained static word embeddings for Portuguese (BR+PT), trained by NILC on a large multi-genre corpus (~1.39B tokens, 17 sources).