Published June 7, 2021 | Version 1.0
Software Open

The Helsinki submission to the AmericasNLP shared task

  • 1. University of Helsinki

Description

Pretrained multilingual translation model (Model B-100dev) for use with OpenNMT-py and SentencePiece.

The example script (translate.sh) shows how to use the model for translating into one of the 10 American languages included in the shared task.

Paper describing our submission: https://www.aclweb.org/anthology/2021.americasnlp-1.29/

GitHub repository with data and training details: https://github.com/Helsinki-NLP/americasnlp2021-st

Files

Files (3.2 GB)

Name Size Download all
md5:a42ddaa0c3703304a13b9d5ed3f30c39
518 Bytes Download
md5:7826132834158935c2433e5848883856
3.2 GB Download
md5:2e8819c6b7060f32f6c98b6f8a936fd0
801.6 kB Download
md5:f5afd188aaa89ad46cf852e78ad324b4
827 Bytes Download

Additional details

Related works

Is referenced by
Software documentation: https://github.com/Helsinki-NLP/americasnlp2021-st (URL)
Is supplement to
Conference paper: https://www.aclweb.org/anthology/2021.americasnlp-1.29 (URL)

Funding

FoTran – Found in Translation – Natural Language Understanding with Cross-Lingual Grounding 771113
European Commission