Urdu Text Reuse Detection

Final Year Project (FYP) | COMSATS University Islamabad, Lahore Campus
Session 2021–2025

---

📚 Project Overview

In today's digital world, the unauthorized reuse and plagiarism of content, especially in low-resource languages like Urdu, has become increasingly common.

Urdu Text Reuse Detection is a research and development project aimed at building Urdu corpora using a semi-automated back-translation approach and applying Machine Learning (ML), Deep Learning (DL), and Large Language Model (LLM) techniques for detecting text reuse and plagiarism.

This project also provides a user-friendly web application and REST APIs for practical usage.

---

🌟 Objectives

• Creation of 9 Urdu Text Reuse Corpora using back-translation techniques.
• Training ML and DL models to predict and classify text reuse:
  • Partially Derived (PD)
  • Wholly Derived (WD)
  • Non-Derived (ND)
• Development of a Web Application for plagiarism and reuse detection.
• Deployment of REST APIs for external integrations.
• Contribution to Urdu NLP Resources.

---

📜 Problem Statement

• Lack of large-scale, publicly available Urdu text reuse datasets.
• Urdu's complex grammar makes reuse detection difficult.
• Limited research in Urdu plagiarism detection.

---

⚙️ Features

• Creation of multiple Urdu corpora from the COUNTER dataset via back-translation.
• Application of ML (SVM, KNN, Naive Bayes) and DL (CNN, LSTM, Transformers).
• Web App with:
  • Login, Registration, Profile Management
  • Urdu Text Reuse and Plagiarism Detection
  • OCR support for image-to-text extraction
  • API Key Management and Usage Analytics
• REST API services with Rate Limiting and Monitoring.
• Admin Dashboard for system and user management.

---

🛠️ Tech Stack

• ML/DL Frameworks: Scikit-learn, TensorFlow, Huggingface Transformers
• Backend: Python (Flask / Django)
• Frontend: React.js / Next.js
• Database: PostgreSQL / MySQL
• OCR: Tesseract OCR
• Authentication: JWT
• Hosting: AWS / Heroku / DigitalOcean

---

🧪 Project Structure

Root Directory
|
|— Web Development App
|     |— public/
|     |— src/
|     |— package.json
|     |— README.md
|
|— Python Server API
|     |— app/
|     |— models/
|     |— static/
|     |— templates/
|     |— requirements.txt
|
|— Native App
      |— android/
      |— ios/
      |— src/
      |— App.js
      |— package.json

• Web Development App: Frontend source code (React.js based)
• Python Server API: Backend server code with API routes, ML models, and database integration.
• Native App: Mobile application (React Native) for mobile reuse detection.

---

Testing

• Unit Testing
• Integration Testing
• Compatibility Testing
• Acceptance Testing

---

🏆 Achievements

• Developed 9 new Urdu reuse corpora via innovative back-translation.
• Integrated trained ML/DL models into a live web system.
• Provided public API access for text reuse detection.
• Contributed to Urdu NLP research and resources.

---

🚀 Future Work

• Expanding datasets with more Urdu sources.
• Cross-language reuse detection (e.g., Urdu-English).
• Model optimization for better accuracy and performance.
• Mobile application enhancement for offline use.

---

✍️ Contributors

• Muhammad Umer Aamir (FA21-BSE-114)
• Malik Ashas (FA21-BSE-120)
• Usama Tufail (FA21-BSE-053)

Supervisor:
Dr. Muhammad Sharjeel (Assistant Professor, CUI Lahore)

---

Note: This project focuses only on the Urdu language. Cross-lingual and unsupervised approaches are not covered.