Sam-Tek/Amharic-Text-Extraction-and-Translation

This project was created to automatically translate the book of Clementos from ethiopian bible # Amharic OCR & Translation Pipeline (GUI Automation) ## 📌 Overview This project was created to **automatically translate the book of Clementos**. The original material was a **scanned PDF** containing **Amharic and Ge'ez (Gueez)** text. Because the document was image-based, no direct text extraction was possible. To solve this, we built a **step-by-step automation pipeline** combining external tools and custom scripts. High-level workflow: 1. **Scanned PDF → cropped images** 2. **Images → Amharic text (OCR)** using Google Translate 3. **Amharic text → English (or other languages)** using ChatGPT --- ## 🧠 Why This Approach? ### ❌ What does NOT work well * ChatGPT: image → Amharic text (OCR) ❌ * Google Translate: image → English translation ❌ (quality is not satisfactory) ### ✅ What works better * Google Translate: image → **Amharic text only** ✅ * ChatGPT: **Amharic text → English (or other language)** ✅ ➡️ This project automates that exact workflow. --- ## 🧩 Project Structure ``` project/ ├── ExtractTextImage.py # Google Translate OCR automation (image → Amharic text) ├── ChatGPTScraper.py # ChatGPT automation (Amharic text → translation) ├── bouton_parcourir.png # UI reference image (Browse button) ├── bouton_copier.png # UI reference image (Copy button) ├── bouton_close.png # UI reference image (Close button) ├── copy.png # UI reference image (ChatGPT copy button) ├── images/ # Folder containing Amharic images ├── outputChatgpt.txt # Translation output └── README.md ``` --- ## 📄 Source Preparation (PDF → Images) The book of Clementos was provided as a **scanned PDF** mixing **Amharic and Ge'ez**. We used **ilovepdf.com** for preprocessing: * Crop pages to remove the Ge'ez text * Convert each PDF page into a separate **JPG image** These images are then used as input for the OCR step. --- ## ⚙️ Requirements * Python 3.9+ * Firefo …