Are the splits accurate?

#3
by janaab - opened

The current splits are:

  • DEV : 216 rows
  • TEST : 232 rows

This is different from the splits (train, dev, test) documented in the Dataset card. Looking at the counts, my best guess is that the current splits were formed as follows - can you confirm:

  • DEV : train + dev
  • TEST : train + dev + test

Hi did you solve this? I think that the README.md might be copied from another dataset and it has not been changed.

If you download the VOXCONVERSE from the original page, it has the same structure.

Sign up or log in to comment