Downloads · 30 days
8
5% of all-time downloads
arabic-fake-news-arbert-afnd
zhafyz/arabic-fake-news-arbert-afnd
arabic-fake-news-arbert-afnd is a text classification model from zhafyz. Use it when you need a label for a piece of text. The card lists the license as apache-2.0.
Fine-tuned UBC-NLP/ARBERT on the AFND dataset (600k+ Arabic news articles). Classifies Arabic news text as credible or notcredible.
All-time downloads
168
Public
Parameters
163M
651 MB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors651 MB · 100%
At a glance
- Task
- Text Classification
- License
- apache-2.0
- Model type
- bert
- Access
- Public
- Created
- May 5, 2026
- Updated
- May 5, 2026
- SHA
- a09ef077
Try a prompt
Datasets
Model card
From the Hugging Face model README
Arabic Fake News Detector — ARBERT (MSA / Formal News)
Fine-tuned UBC-NLP/ARBERT on the AFND dataset (600k+ Arabic news articles).
Classifies Arabic news text as credible or not_credible.
Best for: Formal Modern Standard Arabic (MSA) news articles.
Reference benchmark: 98.8% accuracy (Nassif et al. 2022)
Usage
from transformers import pipeline
import pyarabic.araby as araby, re
def clean(text):
text = re.sub(r'http\S+|<[^>]+>|[@#]', ' ', text)
text = araby.strip_tashkeel(text)
text = araby.strip_tatweel(text)
text = araby.normalize_alef(text)
text = araby.normalize_teh(text)
text = re.sub(r'[^\u0600-\u06FF0-9\s]', ' ', text)
return re.sub(r'\s+', ' ', text).strip()
clf = pipeline('text-classification', model='YOUR_HF_USERNAME/arabic-fake-news-arbert-afnd')
result = clf(clean('نص الخبر العربي هنا...'))
Performance
| Metric | Score |
|---|---|
| Accuracy | ~98% |
| F1 (macro) | ~97% |
References
- Nassif et al. (2022) — Arabic Fake News Detection Based on Deep Contextualized Embedding Models
- Khalil et al. (2022) — AFND: Arabic fake news dataset
- Task
- Text Classification
- Type
- bert
- License
- apache-2.0
- Languages
- ar
- Created
- May 5, 2026
- Updated
- May 5, 2026