Datasets:

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

Which language detector did you use

#28
by ming030890 - opened

The yue classification is clearly wrong unfortunately

Many Japanese text is classified as yue

Thanks

FineData org

Hi, we have used glotlid.
Indeed it's possible, we had similiar issue with fw2 where many of minor languages are missclasified as major language

Thanks for getting back to me. Let me see if I can contribute to glotlid so that we will lucky have a good Cantonese dataset next time

(Btw , Japanese is not even the major language of Cantonese, haha)

Thanks for the great work!

ming030890 changed discussion status to closed

Sorry one more question, do you use V2 or v3 of it?

Just in case it is already fixed in v3

Thanks

FineData org

v3

Sign up or log in to comment