This project focuses on automatically generating Arabic questions from input context using a fine-tuned AraT5 transformer model. It also includes a complete MERN stack web application for user interaction and evaluation.
The main goal is to assist Arabic learners and educators by generating high-quality questions automatically from a given Arabic text. The project supports:
- Natural Language Processing (NLP) for the Arabic language.
- Educational tools through AI-powered question generation.
- React.js
- Tailwind CSS (optional)
- Node.js
- Express.js
- MongoDB (for storing user data and chat history)
- Python
- FastAPI
- HuggingFace Transformers (AraT5 model)
This folder contains all resources related to Arabic question generation and answering using deep learning and NLP techniques. It leverages datasets such as Arabic-SQuAD, ARCD, MLQA, and TydiQA to train and evaluate models for generating high-quality, answerable questions from Arabic context passages.
Structure:
- notebooks/: Jupyter notebooks for experiments, preprocessing, training, and evaluation.
- scripts/: Utility scripts for text preprocessing and testing.
- results/: Output and evaluation reports (e.g., generated questions, model test results).
Refer to the notebooks for step-by-step code, explanations, and example usages. Example context and generated questions are provided in the results files.
- 💬 Input Arabic context and generate multiple questions.
- 🧠 Model fine-tuned on Arabic question generation datasets.
- 💾 Save and access previous chats (context + questions).
# Clone the repo
git clone https://github.com/your-username/arabic-question-generator.git
cd arabic-question-generator
# Install client and server dependencies
cd frontend
npm install
cd backend
npm install
# Start frontend and backend
npm run dev # or use nodemon / dev script
-
For more details and experiments on the question model, visit the dedicated repository: Question Model Repository