Datasets:
Tasks:
Text Classification
Sub-tasks:
multi-class-classification
Languages:
Tagalog
Size:
1K<n<10K
License:
Test data split maybe wrong?
#3
by
selfhwa
- opened
I just wanted to note this here that train/test split of this dataset may not be accurate as reported here: https://github.com/bazingagin/npc_gzip/issues/13#issuecomment-1641129777
Hi @selfhwa , could it be this line that leads to the data leakage? https://huggingface.co/datasets/dengue_filipino/blob/363b2d57046b934880993b14650956b01f20ceb8/dengue_filipino.py#L77
Hi,
Thanks @selfhwa for reporting and thanks @sabilmakbar for your investigation.
Yes, there is a bug in the loading script. This should be fixed by #5.
Fixed by #5.
albertvillanova
changed discussion status to
closed