Hebrew Corpus
This corpus contains offensive language in Hebrew manually annotated. The data includes 15,881 tweets, labeled with one or more of five classes (abusive, hate, violence, pornographic, or non-offensive). The corpus is annonated manually by Arabic-Hebrew bilingual speakers.
Model Download
Huggingface: https://huggingface.co/SinaLab/OffensiveHebrew
Models
Best model (main branch: https://huggingface.co/SinaLab/OffensiveHebrew/tree/main), with micro-F1 score of 0.82
AlephBERT (AlephBERT branch: https://huggingface.co/SinaLab/OffensiveHebrew/tree/AlephBERT) consists of eight models trained on eight different datasets described in the paper.
HeBERT (HeBERT branch: https://huggingface.co/SinaLab/OffensiveHebrew/tree/HeBERT) consists of eight models trained on eight different datasets described in the paper.
Github Repository
git clone https://github.com/SinaLab/OffensiveHebrew
You can download the data from the following GitGub link:
- Downloads last month
- 0