Loading datasets...
Task: MT
Release Date: 5/14/2026
Format: MP3, TSV
Size: 12.56 GB
Share
CoVoST 2 is a large-scale multilingual speech to text translation corpus based on Mozilla Common Voice 4.0. This segment of the corpus contains the English audio (436 hours) and the translations in Arabic.
Licensing
Creative Commons Attribution Non Commercial 4.0 International (CC-BY-NC-4.0)
https://spdx.org/licenses/CC-BY-NC-4.0.htmlRestrictions/Special Constraints
Research and non-commercial use only.
Forbidden Usage
You agree not to attempt to determine the identity of speakers in this dataset. You agree not to train models for public distribution on this dataset. Any attempt to clone the voice or train models that imitate the speakers in this dataset is forbidden.
Ethical Review
This dataset contains data from speakers who have asked to be removed from the Mozilla Common Voice dataset, we expect that you will treat it with care. We expect it to be used only for research and non-commercial purposes only.
Intended Use
Replication experiments involving the CoVoST 2 datasets.
This dataset contains 320495 audio clips totalling 436:04:23 of audio in English with the corresponding translations in Arabic.
End-to-end speech-to-text translation (ST) has recently witnessed an increased interest given its system simplicity, lower inference latency and less compounding errors compared to cascaded ST (i.e. speech recognition + machine translation). End-to-end ST model training, however, is often hampered by the lack of parallel data. Thus, we created CoVoST, a large-scale multilingual ST corpus based on Common Voice, to foster ST research with the largest ever open dataset. Its latest version covers translations from English into 15 languages---Arabic, Catalan, Welsh, German, Estonian, Persian, Indonesian, Japanese, Latvian, Mongolian, Slovenian, Swedish, Tamil, Turkish, Chinese---and from 21 languages into English, including the 15 target languages as well as Spanish, French, Italian, Dutch, Portuguese, Russian. It has total 2,880 hours of speech and is diversified with 78K speakers.
path: Filename of the audio file
sentence: The sentence in the source language
translation: The sentence in the target language
client_id: The ID of the speaker of the source language, used for maintaining hygiene in the splits.
path sentence translation client_id
common_voice_en_19664034.mp3 "These data components in turn serve as the ""building blocks"" of data exchanges." تمثل عناصر البيانات هذه في المقابل "العناصر الأساسية" لتبادل البيانات. 4f29be8fe932d773576dd3df5e111929f4e222422322450983695eaa8625a12659cd3e999a061a29ebe71783833bebdc2d0ec6b97e9a648bf6d28979065f85ad
common_voice_en_19664035.mp3 The church is unrelated to the Jewish political movement of Zionism. الكنيسة ليست على صلة بحركة الصهيونية اليهودية السياسية. 4f29be8fe932d773576dd3df5e111929f4e222422322450983695eaa8625a12659cd3e999a061a29ebe71783833bebdc2d0ec6b97e9a648bf6d28979065f85ad
common_voice_en_19664037.mp3 The following represents architectures which have been utilized at one point or another. يمثل الآتي الهياكل التي تم استخدامها في مرحلة أو أخرى. 4f29be8fe932d773576dd3df5e111929f4e222422322450983695eaa8625a12659cd3e999a061a29ebe71783833bebdc2d0ec6b97e9a648bf6d28979065f85ad
common_voice_en_19664038.mp3 Additionally, the pulse output can be directed through one of three resonator banks. علاوة على أن خرج النبض يمكن توجيهه عبر مصارف الرنان. 4f29be8fe932d773576dd3df5e111929f4e222422322450983695eaa8625a12659cd3e999a061a29ebe71783833bebdc2d0ec6b97e9a648bf6d28979065f85ad
| # clips | |
|---|---|
| Train | 289431 |
| Dev | 15532 |
| Test | 15532 |
If you use this dataset in your work please cite
@misc{wang2020covost,
title={CoVoST 2: A Massively Multilingual Speech-to-Text Translation Corpus},
author={Changhan Wang and Anne Wu and Juan Pino},
year={2020},
eprint={2007.10310},
archivePrefix={arXiv},
primaryClass={cs.CL}
}