Toxic Comment Detector

The Toxic Comment Detector leverages Hugging Face’s unitary/toxic-bert model to identify harmful, offensive, or abusive language in user-generated content. Built with a ReactJS (Vite) frontend and a Flask backend, this tool provides real-time toxicity analysis for enhancing online safety and moderation.

---

Features

• Toxicity Detection: Classifies user-provided text as "Toxic" or "Non-Toxic".
• Detailed Insights: Provides confidence scores for specific toxic traits like hate speech or profanity.
• Real-Time Feedback: Processes and returns results instantly through an intuitive interface.
• Open-Source and Customizable: Easily extendable for additional features or use cases.

---

Tech Stack

Frontend

• ReactJS with Vite for a fast and responsive UI.
• Axios for API communication.

Backend

• Flask for handling API requests and processing text.
• Hugging Face’s unitary/toxic-bert for toxicity detection.

---

Installation and Setup

1. Clone the Repository

git clone [email protected]:allanninal/toxic-comment-detector.git
cd toxic-comment-detector

2. Backend Setup

1. Create and activate a virtual environment:

   python3.12 -m venv venv
   source venv/bin/activate  # Linux/Mac
   venv\Scripts\activate     # Windows

2. Install dependencies using requirements.txt from the backend folder:

   pip install -r backend/requirements.txt

3. Run the Flask backend:

   python backend/app.py

3. Frontend Setup

1. Navigate to the frontend directory:

   cd frontend

2. Install dependencies:

   npm install

3. Start the React development server:

   npm run dev

Visit the app at http://localhost:5173.

---

How It Works

1. Input Text: Enter a comment or text to analyze for toxicity.
2. Backend Processing: Flask processes the input text using Hugging Face’s unitary/toxic-bert model.
3. Display Results: The frontend shows the toxicity classification and confidence scores in real-time.

---

Future Enhancements

1. Multi-Language Support: Integrate multilingual models like xlm-roberta-large to analyze non-English text.
2. Custom Toxicity Categories: Allow fine-tuning models for user-defined toxicity classifications.
3. Explainability: Provide detailed insights into why a comment is classified as toxic.
4. Live Moderation: Integrate real-time toxicity detection with chat systems for instant moderation.

---

Alternatives for Toxic Comment Detection

1. cardiffnlp/twitter-roberta-base-offensive: Best for social media applications like Twitter.
2. facebook/bart-large-mnli: Offers zero-shot classification for custom toxicity categories.
3. Perspective API (Google): Provides multi-dimensional toxicity scores for hate speech, threats, and more.
4. TextBlob with Custom Rules: A lightweight library for sentiment and basic toxicity analysis.

---

License

This project is licensed under the MIT License. See the LICENSE file for details.

---

Support

If you find this project helpful, consider supporting me on Ko-fi:
ko-fi.com/allanninal

---

Explore More Projects

For more exciting projects, check out my list of AI Mini Projects:
Mini AI Projects GitHub List