README.md

metadata

title: Human vs AI Text Detector
emoji: 🔍
colorFrom: purple
colorTo: blue
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: false
license: apache-2.0
tags:
  - text-classification
  - human-ai-text-attribution
  - african-languages
  - multilingual
  - hata

🔍 Human vs AI Text Detector

Detect whether text is human-written or AI-generated across multiple African languages!

🌟 Features

• Multilingual Support: Works with English, Yoruba, Hausa, Igbo, Swahili, Amharic, and Nigerian Pidgin
• High Accuracy: 100% accuracy on validation set
• Fair & Unbiased: Explicitly trained with fairness constraints across all languages
• Easy to Use: Simple interface for single text or batch processing

🎯 How to Use

1. Single Text: Paste your text in the input box and click "Classify Text"
2. Batch Processing: Upload a .txt file with one text per line for batch classification
3. Examples: Try the pre-loaded examples in different languages

🔬 Model Details

• Base Model: AfroXLMR-base
• Parameters: ~270M
• Training: Fine-tuned on PhD HATA African Dataset
• Fairness: EOD = 0.0, AAOD = 0.0 (perfect fairness)

📊 Performance

Metric	Score
Accuracy	100%
F1 Score	100%
Precision	100%
Recall	100%

⚠️ Limitations

• Optimized for African languages in training set
• Performance may vary on newer AI generation systems
• Should be used as part of a broader content verification system

🔗 Links

• Model Repository
• Training Visualizations
• Dataset

📚 Citation

@misc{msmaje2025hata,
  author = {Maje, M.S.},
  title = {AfroXLMR for Human-AI Text Attribution},
  year = {2025},
  publisher = {HuggingFace},
  url = {https://huggingface.co/msmaje/phdhatamodel}
}

📧 Contact

For questions or feedback, please open an issue in the model repository.