Detect cyberbullying in code-mixed social-media text (Romanized Hindi, Hindi, and English) with this easy-to-use deep-learning model. Perfect for websites, software, or mobile apps that need real-time content moderation.
Handles noisy, multilingual code-mixed text with a CNN–BiLSTM model, including full preprocessing, trained from scratch for high accuracy, and ready for integration via Python API, Flask/FastAPI, or TensorFlow Lite.
-Detects abusive, harassment, or bullying content -Fast, lightweight, and adaptable to new datasets -Open-source and easy to modify
-
Clone the repo
-
Install dependencies (tensorflow, keras, numpy, pandas, scikit-learn)
-
Load the pre-trained model and run predictions on your own text
This project is licensed under the Apache License 2.0 — free to use and modify for personal, academic, or commercial projects.