Datasets:

espnet

/

yodas

like 89

I am not sure if I missed this anywhere in the docs.

Could you please confirm my hunch is right that the utt_id is composed by
'some_youtube_video_hash - incremental_int - start_segment_from_video_s -end_segment_from_video_s'

If so, how was the first component generated? It is not a video_id from Youtube right?