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}
}