Daily Paper Atlas
A visualization tool for exploring papers and authors network graphs, inspired by the Model Atlas paper.
Features
- Interactive graph visualization of papers and authors
- Search functionality to find specific nodes
- Color coding by node attributes
- Group filtering
- Node details panel
- Responsive zoom and navigation
Setup
- Convert your GRAPHML file to JSON format:
python graphml_to_json.py graph.graphml paper_atlas_data.json paper_atlas_data.json.gz
- Start the web server:
python app.py
- Open your browser to http://localhost:7860
Technology Stack
- Python for server and data processing
- SigmaJS for graph visualization
- jQuery for DOM manipulation
- Pako.js for client-side decompression
- HTML/CSS for the interface
Data Format
The application expects a GraphML file that will be converted to a JSON format suitable for SigmaJS. The GraphML file should include node attributes like:
- label
- type (e.g., "author", "paper")
- x, y (coordinates for positioning)
- size
- r, g, b (color values)
Credits
This visualization is inspired by the work described in "Charting and Navigating Hugging Face's Model Atlas" by Eliahu Horwitz, Nitzan Kurer, Jonathan Kahana, Liel Amar, and Yedid Hoshen.