Datasets:
Mismatched language labels
#28
by
RishikesshB
- opened
I’ve come across a few cases where the language labels seem to be incorrect. For example, some audios that are actually in Hindi are marked as Tamil, and even the transcriptions are in Hindi.
I initially assumed the audio and transcription would always match, but I’ve also found instances where they don’t. In some cases, the audio is in Hindi while the transcription is in Kannada, with no clear relation to what’s spoken.
Could you please clarify if this is a known issue?