📘 English-to-Persian PDF Translator (Offline)

This project provides a local deep learning pipeline for translating English PDF documents into Persian (Farsi) using MBART multilingual models from Hugging Face. It supports PDF text extraction, OCR for scanned pages, and right-to-left Persian text rendering.

---

🚀 Features

• ✅ Translate English → Persian using MBART-50
• 🧠 Works fully offline (after initial model download)
• 📄 Extracts text from both digital and scanned PDFs
• 🔤 Handles Persian text reshaping and BiDi correction
• 🧩 Automatically installs dependencies and prepares directories
• 📦 Saves outputs (translated text files or PDFs) in outputs/

---

🛠️ Installation

1️⃣ Clone the repository

git clone https://github.com/pouya-mhb/PDF2Persian.git
cd PDF2Persian

2️⃣ Run the setup script

This installs all required packages and prepares the project directories:

python install_local_project.py

It will automatically:

• Install required Python libraries
• Create an outputs/ folder for results
• Download NLTK tokenizer data

---

📚 Usage

1️⃣ Place your English PDF

Put your file (e.g., sample.pdf) in the project folder.

2️⃣ Run the translator

python pdf_translator_local_models.py

This script will:

• Extract and clean text from the PDF
• Translate content to Persian
• Fix Persian text order (using arabic_reshaper + python-bidi)
• Save the translated output to outputs/translated_sample.txt

---

🧩 Directory Structure

pdf-persian-translator/
│
├── install_local_project.py        # Installs dependencies and sets up environment
├── pdf_translator_local_models.py  # Main translation script
├── outputs/                        # Folder for translated outputs
├── models/                         # (Optional) for storing local model weights
└── README.md                       # Project documentation

---

⚙️ Dependencies

Installed automatically by install_local_project.py:

torch
transformers
sentencepiece
pdfplumber
pdf2image
pytesseract
layoutparser
arabic-reshaper
python-bidi
nltk
pillow
opencv-python
requests

---

📝 Notes

Make sure Tesseract OCR is installed on your system if you want to process scanned PDFs.

Persian text is reshaped and rendered correctly using:

from arabic_reshaper import reshape
from bidi. algorithm import get_display

final_text = get_display(reshape(translated_text))

• The model (facebook/mbart-large-50-many-to-many-mmt) will be automatically downloaded the first time you run the script.

---

🧠 Example

Input (English):

This study aimed to investigate the impact of using ChatGPT as a learning tool on students' motivation.

Output (Persian):

این مطالعه با هدف بررسی تأثیر استفاده از ChatGPT به عنوان ابزار یادگیری بر انگیزه دانشجویان انجام شد.

---

📄 License

This project is released under the MIT License. Feel free to use, modify, and distribute with proper credit.