PDFGraph - Multilingual PDF Network Analysis

A Python tool for semantic analysis and interactive visualization of PDF document collections.

Features

• ✅ Automatic text extraction from PDFs
• ✅ Multilingual language detection (German, English, French, Spanish, Italian)
• ✅ Semantic similarity calculation (TF-IDF + Cosine similarity)
• ✅ Automatic document clustering
• ✅ Interactive HTML network visualization
• ✅ Double-click to open PDFs
• ✅ Detailed cluster analysis

Installation

Requirements

• Python 3.7+
• pip

Install Dependencies

pip install -r requirements.txt

requirements.txt Content:

PyPDF2
networkx
matplotlib
scikit-learn
numpy
pyvis
langdetect

Usage With Custom Parameters

from PDFGraph import main

main(
    folder_path="./my_pdfs",
    similarity_threshold=0.25,      # Connection strength (0.0-1.0)
    n_clusters=3,                    # Number of clusters
    multilingual=True,               # Enable multilingual processing
    repulsion_strength=2.5,          # Node repulsion (1.0-5.0)
    spring_k=10.0                    # Overall distance (3.0-15.0)
)

Output

The program generates:

• network_interactive.html - Interactive network visualization
• Console output - Language distribution, cluster analysis, similarity values

HTML Visualization Controls

Action	Description
Double-click	Opens the PDF file
Drag & Drop	Move nodes around
Scroll	Zoom in/out
Hover	Shows similarity value
Multi-select	Select multiple nodes

Parameter Tuning

repulsion_strength (1.0 - 5.0)

• Controls repulsion of weak connections
• Low (1.0): Compact layout
• High (5.0): Strong dispersion

spring_k (3.0 - 15.0)

• Distance between all nodes
• Low (3.0): Compact network
• High (15.0): Widespread distribution

similarity_threshold (0.0 - 1.0)

• Minimum similarity for connections
• Low (0.1): Many connections
• High (0.5): Only strong connections

Color Coding

• Clusters → Different node colors
• Green edges → High similarity (>0.7)
• Blue edges → Medium similarity (>0.5)
• Gray edges → Low similarity (>0.3)

Supported Languages

• 🇩🇪 German
• 🇬🇧 English
• 🇫🇷 French
• 🇪🇸 Spanish
• 🇮🇹 Italian

Multilingual document collections are automatically detected and processed.

Example Workflow

# 1. Copy PDFs to a folder
mkdir ./my_pdfs
cp *.pdf ./my_pdfs/

# 2. Run PDFGraph
python PDFGraph.py

# 3. Open the generated HTML file
network_interactive.html

Troubleshooting

"No PDFs found" → Make sure the PDF folder exists and contains .pdf files

"pyvis not installed" → Run: pip install pyvis

"No text extracted" → The PDF might be scanned/image-based (OCR required)

Performance

• ~5-10 seconds for 10-20 PDFs
• ~30 seconds for 50+ PDFs
• Depends on text volume and CPU

---

Version: 1.0 | Author: Sebastian Meyer | Date: October 2025