Skip to content

amazon-science/isometric-slt

Isometric Spoken Language Translation

This repo contains dataset (test), recipes, and scripts to set up and evaluate an isometric translation tasks.

Isometric SLT Shared Task

At the IWSLT 2022 evaluation campaign we are organizing an isometric spoken language translation task. The task requires participants to submit their system(s) performance evaluation on the publicly available test set (MuST-C), and blind test set curated by the organizers.

Task Evaluation

Following the task evaluation timeline the blind set can be accessed from ./dataset/isometric-mt-test. For evaluation scripts see ./scripts.

For more details, see shared task evaluation and system submission descriptions.

Baselines

For baseline model training and evaluation, see baselines readme.

Submissions

For participating teams and their submissions, see submissions readme.

Use cases

To evaluate the impact of isometric translation, we take automatic dubbing as a case study. Access sample automatically dubbed videos utilizing translations from baseline and systems submitted for the task.

Security

See CONTRIBUTING for more information.

License

This project is licensed under the CC-BY-4.0 License.

Citation

If you participate in the Isometric SLT shared task or make use of the isometric-mt-test set, please cite:

@article{lakew2021isometricmt,
  title={Isometric MT: Neural Machine Translation for Automatic Dubbing},
  author={Lakew, Surafel M and Virkar, Yogesh and Mathur, Prashant and Federico, Marcello},
  journal={arXiv preprint arXiv:2112.08682},
  year={2021}
}
@article{virkar2022onoffscreenpa,
  title={Prosodic alignment for off-screen automatic dubbing},
  author={Virkar, Yogesh and Federico, Marcello and Enyedi, Robert and Barra-Chicote Roberto},
  journal={arXiv preprint arXiv:2204.02530},
  year={2022}
}