Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
parquet
Size:
10K - 100K
Tags:
language-identification
License: