Answer: No
Currently to train with Kaldi we need to create text
, segments
, wav.scp
, utt2spk
, and spk2utt
files.
Can we now prepare these files using Lhotse scripts from Next-gen Kaldi?
We have to know how to prepare data in Lhotse style, then we can convert from Lhotse style to Kaldi style using Lhotse function called lhotse.kaldi.export_to_kaldi(recordings, supervisions, output_dir, map_underscores_to=None, prefix_spk_id=False)
We can prepare files in Lhotse format, then we can export Lhotse format to current Kaldi format.
We can’t just get those 5 needed files from Lhotse scripts.
#7 data prep youtube auto transcript