Linking Back to Common Voice
Hey this work is Fantastic. But to make it more usable it would be great if the sentence_id and client_id from common voice was maintained. At the moment there is no obvious way to link back to common voice. I have actually managed to use transcripts/gender/accent/age to link back to ~75% of the data by matching.
I need to do this because I want to use common voice for training and GLOBE V2 for testing. Related to Spoken Language ID solution.
Is it not possible to fix this and include sentence_id and client_id in GLOBE V2/V3?
Thank you so much for your kind words and for your interest in the GLOBE dataset.
You’re absolutely right — having access to the original sentence_id and client_id from Common Voice would significantly improve usability and traceability. We’re happy to share that in GLOBE V3, we plan to include both of these fields, along with additional metadata such as CER, SNR, and objective MOS scores.
The V3 release is currently in progress and is expected to be completed by May.
Thanks looking forward to it!