Datasets:
Danish VoxPopuli synthetic/weakly labeled data?
#2
by Lauler - opened
Hi!
As far as I was aware, the VoxPopuli dataset does not contain any (manually) labeled/transcribed danish audio.
I'm curious how you produced text labels for the Danish subset of VoxPopuli. Have you force aligned protocols from the EU with the audio, or have you used synthetic labels generated by ASR models?