Feka - Fake News Detection Model 🕵️

A fine-tuned DistilBERT model specifically trained to detect fake news and misinformation.

Quick Start

from transformers import pipeline

# Load the model
classifier = pipeline(
    "text-classification", 
    model="mohammed-hoode/feka",
    return_all_scores=True
)

# Test it
texts = [
    "Breaking: Scientists discover miracle cure that doctors hate!",
    "The Federal Reserve announced a 0.25% interest rate increase."
]

for text in texts:
    result = classifier(text)
    print(f"Text: {text}")
    
    for score in result:
        label = "Real News" if score['label'] == 'LABEL_0' else "Fake News"
        print(f"{label}: {score['score']:.3f}")
    print("-" * 50)

Model Details

• Base Model: DistilBERT
• Task: Binary Text Classification
• Labels:
  • LABEL_0: Real/Legitimate News
  • LABEL_1: Fake/Suspicious News
• Language: English
• Training: Fine-tuned on diverse news dataset

Usage in Applications

Perfect for:

• News verification systems
• Social media content filtering
• Misinformation detection pipelines
• Educational fact-checking tools

Performance Notes

• Works best with news-style text
• Optimized for English language content
• Should be combined with human oversight
• Performance varies across different domains

Limitations

• Not 100% accurate - use as assistance tool
• May have biases from training data
• Best used with additional verification methods

Built with ❤️ for fighting misinformation.