serdarcaglar/kiraat

KIRAAT — A Turkish Read-Speech Corpus A sentence-aligned read-speech corpus built from publicly available recordings on Turkish audiobook YouTube channels. The channel credits are in the table at the end of this card; every clip carries the channel it came from in the channel column. clips 1,840,404 duration 3,105.7 hours recommended subset 1,547,494 clips / 2,575.2 hours channels 27 speakers (clustered) 90 source recordings 2,680 words (ASR) 21,695,774… See the full description on the dataset page: https://huggingface.co/datasets/serdarcaglar/kiraat.