Tools and resources for open translation services
- based on Marian-NMT
- trained on OPUS data using OPUS-MT-train (New: leaderboard)
- mainly SentencePiece-based segmentation
- mostly trained with guided alignment based on eflomal wordalignments
- pre-trained downloadable translation models (matrix view), CC-BY 4.0 license
- more freely available translation models from the Tatoeba translation challenge, CC-BY 4.0 license
- demo translation interface available from https://opusmt.wmflabs.org/