NILC Portuguese Word Embeddings — FastText Skip-Gram 600d

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

This repository contains the FastText Skip-Gram 600d model in safetensors format.

📂 Files

embeddings.safetensors → word vectors ([vocab_size, 600])
vocab.txt → vocabulary (one token per line, aligned with rows)

🚀 Usage

from safetensors.numpy import load_file

data = load_file("embeddings.safetensors")
vectors = data["embeddings"]

with open("vocab.txt") as f:
    vocab = [w.strip() for w in f]

word2idx = {w: i for i, w in enumerate(vocab)}
print(vectors[word2idx["rei"]])  # vector for "rei"

Or in PyTorch:

from safetensors.torch import load_file
tensors = load_file("embeddings.safetensors")
vectors = tensors["embeddings"]  # torch.Tensor

📖 Reference

@inproceedings{hartmann-etal-2017-portuguese,
  title        = {{P}ortuguese Word Embeddings: Evaluating on Word Analogies and Natural Language Tasks},
  author       = {Hartmann, Nathan  and Fonseca, Erick  and Shulby, Christopher  and Treviso, Marcos  and Silva, J{'e}ssica  and Alu{'i}sio, Sandra},
  year         = 2017,
  month        = oct,
  booktitle    = {Proceedings of the 11th {B}razilian Symposium in Information and Human Language Technology},
  publisher    = {Sociedade Brasileira de Computa{\c{c}}{\~a}o},
  address      = {Uberl{\^a}ndia, Brazil},
  pages        = {122--131},
  url          = {https://aclanthology.org/W17-6615/},
  editor       = {Paetzold, Gustavo Henrique  and Pinheiro, Vl{'a}dia}
}

📜 License

Creative Commons Attribution 4.0 International (CC BY 4.0)