Dataset · Overfit-GM
77.8K Turkish texts with toxicity labels
Overfit-GM/turkish-toxic-language
Combined corpus of Turkish messages with binary offensive content labels for moderation.
What's inside
The Turkish text dataset includes 77,800 examples combined from several open sources. Each entry is labeled as offensive or non-offensive, with the authors using an ensemble of transformer models to generate pseudo-labels for completeness.
What you can do with it
- Training toxicity classifiers and moderation filters for the Turkish language
- Testing generative model safety with Turkish text
- Tasks
- Text classification
- Languages
- Turkish
- Size
- 10K–100K rows
- Downloads this month
- 199
- Likes
- 36
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("Overfit-GM/turkish-toxic-language")