📄 Paper Anonymizer

A web-based tool to anonymize research papers by removing author names, affiliations, and other identifying information before peer review.

---

🚀 Features

• 📂 Select input and output folders
• 📄 Supports PDF, DOC, DOCX files
• 🔄 Automatic DOC/DOCX → PDF conversion using Microsoft Word COM
• 🖱️ Interactive PDF viewer with drag-to-select redaction
• ✂️ Remove selected regions precisely (word-level redaction)
• 👁️ Preview removals before saving
• ↩️ Undo applied redactions
• 💾 Save anonymized files to output folder
• 🔗 Merge acknowledgement document (optional)
• 🧹 Removes PDF metadata for full anonymization

---

🏗️ Project Structure

Anonymizer-Paper/
│
├── backend/
│   ├── app.py              # FastAPI backend
│   ├── converter.py        # DOC/DOCX → PDF conversion (Word COM)
│   ├── pdf_editor.py       # Redaction + metadata removal
│   ├── utils.py            # File utilities
│   ├── requirements.txt    # Dependencies
│   └── temp/               # Temporary converted PDFs
│
├── input/                  # Input papers
├── output/                 # Anonymized papers
│
├── app.js                  # Frontend logic
├── index.html              # UI
├── style.css               # Styling

.
├── app.py              # FastAPI backend
├── converter.py        # DOC/DOCX → PDF conversion (Word COM)
├── pdf_editor.py       # Redaction + metadata removal
├── utils.py            # File utilities
├── requirements.txt    # Dependencies
│
├── index.html          # Frontend UI
├── app.js              # Frontend logic
├── style.css           # UI styling
│
├── input/              # Input papers
├── output/             # Anonymized papers
├── temp/               # Temporary converted PDFs

---

⚙️ Setup Instructions

1. Clone / Download

git clone <repo-url>
cd paper-anonymizer

---

2. Create Virtual Environment

python -m venv .venv
.venv\Scripts\activate   # Windows

---

3. Install Dependencies

pip install -r requirements.txt

Dependencies include:

• FastAPI
• PyMuPDF
• pywin32 (for Word conversion)

---

4. Enable Word COM (IMPORTANT)

python -m win32com.client.makepy

Then select:

Microsoft Word XX.X Object Library

⚠️ Requires:

• Windows OS
• Microsoft Word installed

---

5. Run Backend Server

uvicorn app:app --reload

Server runs at:

http://localhost:8000

---

6. Open Frontend

Open:

index.html

Or run via Live Server:

http://localhost:63342/.../index.html

---

🧠 How It Works

1. File Loading

• Backend lists files using list_files()
• DOC/DOCX files are converted using Word COM

2. PDF Rendering

• Uses PDF.js to render pages in browser
• Text layer enables accurate selection

3. Selection System

• User drags to select regions
• Coordinates are converted to PDF space

4. Redaction Engine

From pdf_editor.py:

• Extracts words using:

  page.get_text("words")

• Removes only words intersecting selection

• Uses overlap threshold (>20%) for accuracy

5. Metadata Removal

doc.set_metadata({})
doc.del_xml_metadata()

---

🧪 Workflow

1. Select input & output folders
2. Choose a paper from sidebar
3. Drag to select author/affiliation area
4. Click REMOVE (preview)
5. Click SAVE to finalize
6. (Optional) Upload acknowledgement and click MERGE & SAVE

---

⚠️ Known Limitations

• Word → PDF conversion may alter text positioning
• Complex layouts (multi-column, tables) may cause slight inaccuracies
• Requires Microsoft Word (not cross-platform)

---

🔮 Future Improvements

• 🤖 Auto-detect author sections
• 📊 Confidence score for anonymization
• 🧠 NLP-based entity removal (names, emails, institutions)

---

🛠️ Tech Stack

Frontend:

• HTML, CSS, JavaScript
• PDF.js

Backend:

• FastAPI
• PyMuPDF (fitz)
• pywin32 (Word COM)

---

📌 Notes

• Output files are saved as:

  <original_name>_anonymized.pdf
  

• Temporary files stored in /temp

• Supports recursive folder scanning

---

📸 Screenshots

---

⭐ Summary

This tool provides a semi-automated anonymization pipeline combining:

• manual precision (user selection)
• automated processing (word-level redaction + metadata removal)

Designed for research paper review workflows where bias-free evaluation is required.