deHATEr

A browser extension designed to detect hate speech from messages on social media platforms. It utilizes a trained AI model called 4j3k.

This is a group project for Artificial Intelligence.

Requirements

• The trained model “wangchanberta-hatespeech” is mandatory.
• Place the model in the directory /deHATEr/models/.

Usage

1. (Optional) (Recommended) Environment Setup (only for the first run)

• Ensure Python 3.9 or higher is installed.
• Navigate to the /deHATEr directory using the terminal.
• Create and activate a virtual environment using the command: python -m venv venv && source venv/bin/activate.
• Install the dependencies using the command: pip install -r requirements.txt.

If you’re using Windows and want GPU acceleration, keep the optional torch-directml dependency. On other platforms, it’s safe to ignore it if the installation fails.

2. Run the API Server

• Navigate back to the main directory /.
• Host the API server using the terminal command: ./start_api_server.sh.

3. Install the deHATEr Extension on Your Browser

• Open Chrome and go to chrome://extensions.
• Enable Developer Mode and click on “Load unpacked”.
• Choose the UIExtension folder (the one containing manifest.json). The extension should appear.

Remember, the API server is required to be running while using the extension.

Model Credit

• L. Lowphansirikul, C. Polpanumas, N. Jantrakulchai, and S. Nutanong, “WangchanBERTa: Pretraining transformer-based Thai language models,” arXiv preprint arXiv:2101.09635, 2021. [Online]. Available: https://arxiv.org/abs/2101.09635.
• ForJustice3K, “4j3k: pretrained model for hatespeech detection”, 2025. [Online]. Available: https://github.com/psrisuphan/4j3k.