Text Mining Orkut’s Community Data with Python: Cultural Memory, Platform Neglect, and Digital Amnesia

---

PyData Global 2025

Time: 11:00–11:30 EST | Date: Dec. 9, 2025 | Location: https://pydata.org/global2025

Objective

This repository contains the full code, data, and presentation materials for a study of 124,988 recovered Orkut community names. The project reconstructs a small surviving portion of Orkut’s deleted communities and analyzes them using Python-based stylometry, semantic embeddings, and clustering. The goal is to surface linguistic patterns, cultural themes, and expressive registers from early Brazilian internet culture, and to demonstrate how computational methods can help recover and interpret fragments of lost digital platforms.

Tools Used

• BeautifulSoup (HTML parsing)
• Pandas (data processing)
• SentenceTransformers (multilingual embeddings)
• UMAP, HDBSCAN, KMeans (clustering + projection)
• BERTopic (topic modeling)
• Plotly (visualization)
• Positron IDE (Python editing + integrated terminal)
• Data Explorer (CSV inspection, previews, summaries)
• Quarto Previewer (live slides + reproducible report)

Repository Structure

.
├── README.md                   # Project overview, instructions, talk abstract
├── _quarto.yml                 # Quarto project configuration
├── requirements.txt            # Python dependencies for reproducibility
│
├── orkut_analysis.qmd          # Full analysis: scraping, stylometry, embeddings, clustering
├── orkut-presentation.qmd      # PyData Global 2025 talk slides (Reveal.js)
├── orkut_scraper.py            # Wayback Machine HTML scraper using requests + BeautifulSoup
├── orkut_communities.csv       # Cleaned dataset of 124,988 community names
├── pink.scss                   # Custom SCSS theme for slides (Orkut-inspired design)
│
└── images/                     # All images, figures, and plot outputs
    ├── orkut-logo.png          # Orkut logo for slides + README
    ├── orkut-profile1.jpg      # Early Orkut profile screenshot
    ├── orkut-profile2.jpeg     # Later Orkut profile (friend-provided)
    ├── orkut-wayback.png       # Wayback Machine interface screenshot
    ├── rodrigo-about_me.jpg    # Slide photo
    │
    ├── plot1.png               # Language distribution
    ├── plot2.png               # UMAP semantic map
    ├── plot3.png               # Caps lock stylometry
    ├── plot4.png               # Miguxês stylometry
    ├── plot5.png               # BERTopic micro-topic bars
    ├── plot6.png               # BERTopic intertopic map
    ├── plot7.png               # Additional analysis plot
    ├── plot8.png               # Additional analysis plot
    ├── plot9.png               # Additional analysis plot
    └── plot10.png              # Sentiment distribution map

Usage

Install dependencies:

pip install -r requirements.txt

Render the analysis:

quarto render orkut_analysis.qmd

Preview the presentation slides:

quarto preview orkut-presentation.qmd

Disclaimers

By accessing or using this repository, you agree to the following terms:

• The materials in this repository are provided on an “as is” and “as available” basis for research and educational purposes only.
• No guarantees are made regarding accuracy, completeness, or suitability for any specific use.
• By using this repository, you agree that all use is at your own risk, and that the author assumes no responsibility or liability for any damages, losses, or legal issues arising from your use, modification, or distribution of this code or data.
• Users are solely responsible for ensuring that their use of the included code and data complies with all applicable laws, platform terms of service, and ethical guidelines.
• Nothing in this repository should be interpreted as legal advice.

Contact

Rodrigo Silva Ferreira
rodrigosf672@gmail.com
LinkedIn | GitHub