AI-Text-Detector

🤖 AI Text Detector

This is a web application that analyzes whether a piece of text is AI-generated or human-written using a combination of deep learning models. The app is built with Streamlit and powered by:

• 🧠 RoBERTa-based classifier (roberta-base-openai-detector) for AI text detection
• 📉 GPT-2 language model for perplexity scoring
• ⚙️ PyTorch for model inference
• 🎨 Aesthetic and responsive UI with Streamlit custom styling

---

🚀 Features

• Detects AI-generated vs human-written text
• Provides:
  • Classifier confidence score
  • GPT-2 perplexity (language naturalness)
  • Combined AI-likelihood score
• Clean, interactive web interface
• Works on CPU and GPU

---

📦 Installation

1. Clone the repo:

   git clone https://github.com/yourusername/ai-text-detector.git
   cd ai-text-detector

Set up a virtual environment (optional but recommended):

bash Copy Edit python3 -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate Install dependencies:

bash Copy Edit pip install -r requirements.txt 🧪 Run the App Locally bash Copy Edit streamlit run ai_detector_web.py Then visit: http://localhost:8501

🛠️ Tech Stack Python 3.8+

Streamlit

PyTorch

Transformers (Hugging Face)

GPT-2 (gpt2)

RoBERTa (roberta-base-openai-detector)

🧩 How It Works RoBERTa Classifier:

Returns a confidence score that the input was written by AI.

GPT-2 Perplexity:

Measures how “surprising” the text is to a language model.

Low perplexity = likely AI-generated.

Heuristic Score:

Combines classifier confidence and GPT-2 perplexity to provide a more reliable estimate.