Repository navigation
The first release of the Hate Speech Detector web application. This release includes the initial implementation of the backend, frontend, and core features for identifying and classifying hate speech texts.
Features
- Text Preprocessing Pipeline: Automatically cleans and prepares raw text data for model training and prediction.
- Machine Learning Model: A scikit-learn-based classifier trained to distinguish between toxic and non-toxic text.
- FastAPI Backend: Provides a REST API for predictions and integrates seamlessly with the frontend.
- Next.js Frontend: User-friendly interface for inputting text and viewing classification results in real time.
- Dataset Scripts: Utilities for preprocessing raw datasets and training the machine learning model.