which dataset you would most recommend using?

#2
by Jeremy1110 - opened

Hello, author. I am currently working on multilingual ASR, which includes Korean. I noticed you have provided several datasets, such as KsponSpeech, KrespSpeech, KoreaSpeech, and GyeongsangSpeech. I would like to ask which dataset you would most recommend using?

@Jeremy1110 First of all, thank you for your interest in Korean.

jp1924/KrespSpeech
jp1924/KsponSpeech
jp1924/KconfSpeech
jp1924/MeetingSpeech
This is recommended. The ones with names like GyeongsangSpeech are made up of dialects.

And if you're curious about how the data was created, you can find it here. https://github.com/jp1924/HF_builders

@jp1924

Thank you for your reply. I will try using these datasets for training. Thank you very much!

Jeremy1110 changed discussion status to closed

Sign up or log in to comment