Multi-Lingual AI Text Detection: Bangla and English Classifier

• Tech Stack: PyTorch, Hugging Face Transformers, Streamlit
• Github URL: Project Link
• Website URL: Project Link

Developed Multi-Lingual AI Text Detection, a transformer-based system to classify text as AI-generated or human-written in both Bangla and English.

Fine-tuned the XLM-RoBERTa model (270M+ parameters) on 6,746 samples — 1,746 Bangla and 5,000 English — using stratified splitting to preserve language and label balance, achieving 83.5% accuracy.

Deployed the model as an interactive Streamlit web application on Hugging Face Spaces for real-time text classification with confidence scores.