Transcription langauge is different from audio language
Hi. Thank you so much for open sourcing your dataset!
I found that in the English dataset. There were some cases where the audio was in Chinese but transcriptionw as in English.
I believe there is a mix of dataset for both transcription and translation.
Hi @leduckhai
Can u have a look at this when you are able pls?
Hi,
Thanks for your feedback!
I just quickly checked the dataset again. I did not see any mismatch between transcription and audio as you said.
I guess there were some displaying errors in your browser, because as I saw in your image here, the English dataset has 1k56 rows. Actually the English should have 33.1k rows, while the Chinese dataset has 1k56 rows. That's why you were openning the English dataset but the HuggingFace system showed the Chinese instead, I guess.
Maybe you can try reloading!
Hope it helps
hmm thanks for the feedback.
I guess this might be an error from browser