Project Structure

• ouvia/ – Flask web application that powers the study portal. Handles user authentication, role-based dashboards, conversation entry management, assignment workflows, and the recording/annotation pipeline.
• study/ – Analysis and result-generation code for the paper. Contains data preparation scripts, statistical analyses, quality checks, and automatic translation quality evaluation.

[!NOTE] We will soon release a basic documentation in the GitHub wiki as a tutorial walkthrough for running a new study.

[!WARNING] This is a research project. All results reported in our paper and new studies can be reproduced with the code released here, but this is not a production-ready platform. Support is provided on a best-effort basis — if anything is unclear, please open an issue and we'll do our best to help.

Data Release

We release all the study artifacts, including raw audio recordings, machine translation outputs, automatic quality metrics, and human survey responses collected during the study on Zenodo.

License

The code in this repository is licensed under Apache 2.0. All data artifacts are released under CC BY 4.0.

Citation

@article{attanasio2026ouvia,
  title={Ouvia: A User-centered Framework for Measuring Usability of Speech Translation in Real-World Communication Scenarios},
  author={Attanasio, Giuseppe and Savoldi, Beatrice and Chechelnitsky, Daniel and Negri, Matteo and Carpuat, Marine and Sap, Maarten and Martins, Andr{\'e} FT},
  journal={arXiv preprint arXiv:2606.06177},
  year={2026}
}