mosesdaudu/hausa-bias-detector

hausa Bias Detector

Model Description

This model detects gender bias in hausa text. It classifies text into two categories:

• Gender: Text contains gender bias or stereotypes
• NoBias: Text is neutral and does not contain bias

Intended Use

This model is designed to:

• Detect gender bias in hausa content
• Support content moderation and filtering
• Raise awareness about biased language
• Assist in creating more inclusive hausa content

Training Data

The model was trained on a curated dataset of hausa sentences labeled for gender bias.

Model Performance

• Accuracy: See evaluation metrics in training logs
• F1 Score: See evaluation metrics in training logs

Usage

from transformers import pipeline

# Load the model
detector = pipeline('text-classification', model='mosesdaudu/hausa-bias-detector')

# Detect bias
result = detector("Mwanaume ni mkubwa kuliko mwanamke")
print(result)
# Output: [{'label': 'Gender', 'score': 0.95}]

result = detector("Hakuna tofauti kati ya wanaume na wanawake")
print(result)
# Output: [{'label': 'NoBias', 'score': 0.92}]

Limitations

• The model is trained specifically for hausa language
• Performance may vary on domains not represented in training data
• May not detect all forms of bias
• Should be used as a tool to assist human judgment, not replace it

Ethical Considerations

This model is designed to detect bias, but like all ML models, it may have its own biases. Users should:

• Use this as one tool among many for bias detection
• Validate results with human review
• Be aware of potential false positives/negatives
• Consider cultural context when interpreting results

Citation

If you use this model, please cite:

@misc{hausa-bias-detector,
  author = {Moses Daudu},
  title = {hausa Bias Detector},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/mosesdaudu/hausa-bias-detector}
}

Contact

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