🛡️ AI-Based Comment Moderation System

An end-to-end AI-powered system that analyzes user comments and determines whether they are safe to be displayed. The model performs multi-label toxicity classification and returns probability scores instead of simple binary outputs.

---

🚀 Features

• Multi-label toxicity classification (toxic, obscene, threat, insult, identity hate)
• Probability-based scoring instead of yes/no output
• Real-time moderation (blur / warning for unsafe comments)
• Admin dashboard to review flagged content
• Clean and simple UI for interaction

---

🧠 Model Details

• Algorithm: Logistic Regression
• Vectorization: TF-IDF
• Dataset: Jigsaw Toxic Comment Dataset
• Approach: Multi-label classification

---

🛠️ Tech Stack

Backend: FastAPI (Python) Frontend: HTML, CSS, JavaScript Machine Learning: Scikit-learn Database: Neon (PostgreSQL)

---

🌐 Deployment

• Backend: Render
• Frontend: Vercel
• Database: Neon

---

📸 Screenshots

1. Create Post
   

2. Analysis
   

3. Confirmation
   

4. Feed
   

5. User History
   

6. Admin Dashboard
   

---

🔗 Live Demo

https://moderation-system-deploy.vercel.app/

---

⚙️ How It Works

1. User submits a comment
2. Text is preprocessed and vectorized using TF-IDF
3. Logistic Regression model predicts probabilities for each toxicity label
4. System evaluates risk score
5. UI responds accordingly (normal display / warning / blur)
6. Flagged comments are available in the admin dashboard

---

📦 Setup Instructions

1. Clone the repository

git clone https://github.com/your-username/your-repo.git
cd your-repo

2. Backend setup

pip install -r requirements.txt
uvicorn main:app --reload

Backend will run on:

http://127.0.0.1:8000

---

3. Frontend setup

Open the frontend folder and run using Live Server or any static server.

---

⚠️ Important (Local Development)

When running the project locally, make sure your frontend API calls point to the local backend endpoint instead of the deployed URL.

Example:

// Local
http://127.0.0.1:8000/predict

// Production
https://comment-moderation-api-yx0y.onrender.com/predict

Update your API URLs accordingly depending on your environment.

---

📌 Future Improvements

• Improve model accuracy using deep learning (LSTM / Transformers)
• Add user authentication
• Support multilingual comments
• Improve UI/UX

---

🤝 Contributing

Contributions are welcome! Feel free to fork the repository and submit a pull request.

---

⭐ Acknowledgements

• Jigsaw Toxic Comment Dataset

---

📬 Contact

For feedback or suggestions, feel free to reach out: chaitanyakulkarni1345@gmail.com