Table 5 Applied multilingual dataset separation and hyperparameters.

From: Adaptive ensemble techniques leveraging BERT based models for multilingual hate speech detection in Korean and english

Language/Dataset

Train/valid/test

Batch

Epoch

Learning rate

Chinese

COLD37

20 K train dataset

6 K valid dataset

5 K test dataset

256

10

1e-5

Portuguese

ToLD-Br38

14 K train dataset

3 K valid dataset

4 K test dataset

256

3

3e-6

Back to article page