Difference between revisions of "ASR-nsfc-data"

From cslt Wiki
Jump to: navigation, search
Line 8: Line 8:
  
 
==Uyghur==
 
==Uyghur==
In the second phase, the Uyghur dataset is consist by:
+
In the second phase, the Uyghur dataset consists of:
* 136h speech audio and 353 speakers involved in it.(166 males and 187 females)
+
* 136h speech audio and 353 speakers (166 males and 187 females).
 
* transcription of the speech audio.
 
* transcription of the speech audio.
 
* lexicon in word level.
 
* lexicon in word level.
Line 16: Line 16:
  
 
==Kazakh==
 
==Kazakh==
In the second phase, the Kazakh dataset is consist by:
+
In the second phase, the Kazakh dataset consists of:
* 78h speech audio and 86 speakers involved in it.(40 males and 46 females)
+
* 78h speech audio and 86 speakers (40 males and 46 females).
 
* transcription of the speech audio.
 
* transcription of the speech audio.
 
* lexicon in word level.
 
* lexicon in word level.
Line 24: Line 24:
  
 
==Tibetan==
 
==Tibetan==
In the second phase, the Tibetan dataset is consist by:
+
In the second phase, the Tibetan dataset consists of:
* 72h speech audio and 147 speakers involved in it.(66 males and 81 females)
+
* 72h speech audio and 147 speakers (66 males and 81 females).
 
* transcription of the speech audio.  
 
* transcription of the speech audio.  
 
* lexicon in word level.
 
* lexicon in word level.

Revision as of 02:49, 3 June 2020

Data resources

In order to promote the development of minority speech signal processing technology, we will publish all the M2ASR dataset to scientific research institutions for free. You should ask for license before you can download the dataset.

Please send Email to shiying@cslt.org or lilt@cslt.org to get the license.

Uyghur

In the second phase, the Uyghur dataset consists of:

  • 136h speech audio and 353 speakers (166 males and 187 females).
  • transcription of the speech audio.
  • lexicon in word level.

Download link

Kazakh

In the second phase, the Kazakh dataset consists of:

  • 78h speech audio and 86 speakers (40 males and 46 females).
  • transcription of the speech audio.
  • lexicon in word level.

Download link

Tibetan

In the second phase, the Tibetan dataset consists of:

  • 72h speech audio and 147 speakers (66 males and 81 females).
  • transcription of the speech audio.
  • lexicon in word level.

Download link

Mongolian

Coming soon...

Kirgiz

Coming soon...