Add CHiME4 dataset

[patrickvonplaten]

Adding a Dataset

• Name: Chime4
• Description: Chime4 is a dataset for automatic speech recognition. It is especially useful for evaluating models in a noisy environment and for multi-channel ASR
• Paper: Dataset comes from a channel: http://spandh.dcs.shef.ac.uk/chime_challenge/CHiME4/ . Results paper:
• Data: http://spandh.dcs.shef.ac.uk/chime_challenge/CHiME4/download.html
• Motivation: So far there are very little datasets for speech in datasets. Only lbirispeech_asr so far.

If interested in tackling this issue, feel free to tag @patrickvonplaten

Instructions to add a new dataset can be found here.

[KossaiSbai]

@patrickvonplaten not sure whether it is still needed, but willing to tackle this issue

[KossaiSbai]

Hey @patrickvonplaten, I have managed to download the zip on here and successfully uploaded all the files on a hugging face dataset:

https://huggingface.co/datasets/ksbai123/Chime4

However I am getting this error when trying to use the dataset viewer:

Can you take a look and let me know if I have missed any files please

[KossaiSbai]

@patrickvonplaten ?

[albertvillanova]

Hi @KossaiSbai,

Thanks for your contribution.

As the issue is not strictly related to the datasets library, but to the specific implementation of the CHiME4 dataset, I have opened an issue in the Discussion tab of the dataset: https://huggingface.co/datasets/ksbai123/Chime4/discussions/2 Let's continue the discussion there!