The Centre for Speech Technology Research, The university of Edinburgh

iFlytek Co., Ltd. release of audio recordings for Blizzard 2021

These data are only available to registered participants in the Blizzard Challenge 2021.

You should only request these data after your registration for the challenge has been accepted. Other requests will be ignored.

Speech data

This data is released under a license for non-commercial use only.

Information about the transcriptions

The Spanish data for the hub task 2021-SH1 is provided with text transcriptions only.

For the spoke task 2021-SS1, the training data are the same as for task 2021-SH1. Ten natural recordings of Spanish sentences containing a small numer of English words are also provided, with text transcriptions

Downloads

The following files are available to download after completing the corresponding license. Once we have received your Blizzard Challenge registration and your license, we will email you a password. All requests are manually checked. If you already downloaded spanish_blizzard_release_2021_v1.zip then you can just download spanish_blizzard_release_2021_v2_text_only.zip which provides updated versions of train_text.txt and dev_text.txt.

md5 checksum for the files are as follows:

7a4ad4fa93b7773b1cc1249a397f15dc spanish_blizzard_release_2021_v1.zip (do not use; superseded by v2)
29bc8486f0b61687a58b4bd31e51f176 spanish_blizzard_release_2021_v2.zip
7279191a7e11a0705c19cf5a5494e91c spanish_blizzard_release_2021_v2_text_only.zip

and the sizes of the files are:
spanish_blizzard_release_2021_v1.zip 1.8G
spanish_blizzard_release_2021_v2.zip 1.8G
spanish_blizzard_release_2021_v2_text_only.zip 132K


Contact Simon King for more details.