Repository for sharing the data in the Tamasheq language, one of the languages for the low-resource speech translation track at IWSLT 2022.

This corpus is a collection of radio recordings from Studio Kalangou translated to French. It comprises 17 hours of clean speech in Tamasheq, translated to the French language (taq_fra_clean). We also make available a 19 hour version of this corpus, including 2 additional hours of data that was labeled by annotators as potentially noisy (taq_fra_full). Both versions of this dataset share the same validation and test sets.

In addition to the 17 hours of Tamasheq audio data aligned to French translations, and in light of recent work in self-supervised models for speech processing, we also provide participants with unlabaled raw audio data in the Tamasheq language, as well as in other 4 languages spoken from Niger: French (116 hours), Fulfulde (114 hours), Hausa (105 hours), Tamasheq (234 hours) and Zarma (100 hours). All this data comes from the radio broadcastings of Studio Kalangou and Studio Tamani.

The participants are not required to use any of this data.

Due to its size, this data is made available through this website.

Our paper is available on arXiv.

@article{boito2022speech,
  title={Speech Resources in the Tamasheq Language},
  author={Boito, Marcely Zanon and Bougares, Fethi and Barbier, Florentin and Gahbiche, 
          Souhir and Barrault, Lo\"ic and Rouvier, Mickael and Est\'eve, Yannick},
  journal={Language Resources and Evaluation Conference (LREC)},
  year={2022}
}

All audio recordings are property of Studio Kanlagou, Studio Tamani and Fondation Hirondelle.

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.

This work was funded by the French Research Agency (ANR) through the ON-TRAC project under contract number ANR-18-CE23-0021. This paper was also partially funded by the European Commission through the SELMA project under grant number 957017.