Datasets:

Modalities:
Text
Formats:
parquet
Libraries:
Datasets
Dask
License:

VoxVietnam vs Vietnam-Celeb

#2
by tridm - opened

I've read two articles about your datasets, VoxVietnam and Vietnam-Celeb. My question is: is VoxVietnam an entirely new and separate dataset, or does it contain IDs that already exist in your previously built Vietnam-Celeb dataset?

HuSTeP - Human Speech and Text Processing Lab org

Hi Tri,
Thank you for your question. VoxVietnam MAY have duplicated IDs with Vietnam-Celeb, since they are crawled in different fashions. VoxVietnam does not have speaker identities so we cannot check. We computed the similarity score to exclude IDs in VoxVietnam's training set that may exist in Vietnam-Celeb's test set.

Sign up or log in to comment