Datasets:
Tasks:
Text Classification
Formats:
parquet
Sub-tasks:
sentiment-classification
Languages:
Korean
Size:
100K - 1M
License: