Datasets:
Tasks:
Automatic Speech Recognition
Formats:
parquet
Languages:
Standard Moroccan Tamazight
Size:
1K - 10K
License:
Kabyle Amazigh dataset
#2
by abdelhaqueidali - opened
This is not Moroccan Amazigh, these seem to be the dataset exported from tateba, from the language "Berber".
Though Tatoeba has Kabyle, but most of the data of Berber is in Kabyle as well, and so do the audios.
So basically, Berber on tatoeba are for Kabyle as well. The ones with audios are mostly Kabyle if not all. For other sentences, minority, are in other varinats like Tashelhit.
I can confirm. I've used Tatoeba audio data in the past for ASR, it was 100% Kabyle as far as I can tell.