Experiments on AISHELL-4

[hbredin]

@desh2608 asked about diarization results on AISHELL-4. Thought I'd open a dedicated discussion on pyannote repo.

@hbredin Did you got around to running some diarization baselines on this data?

Still on my TODO list. Wanna try?

I ran VBx and spectral clustering and got the following results:
Method MS FA Conf. DER
VBx 17.06 1 3.97 22.03
Spectral 17.06 1 3.11 21.17

I was wondering if you have any overlap detectors trained on Mandarin?

Did you try pyannote/segmentation with the overlapped speech detection pipeline?
Should work slightly better than goold old ovl-dihard and ovl-ami.

Fine-tuning pyannote/segmentation on AISHELL-4 would definitely help, though.

[desh2608]

Thanks for the suggestions! I'll try the segmentation pipeline and see if it helps. Will get back with updates soon hopefully.