Epstein Files Visualizations

Interactive data visualizations from the November 2025 House Oversight Committee document release

Note: These visualizations are generated from OCR'd documents. Some are verified law enforcement records; others are Epstein's own promotional materials. Cross-reference all findings with independent sources.

Embedding Cluster Map

UMAP dimensionality reduction of all 69,290 document embeddings. Shows semantic similarity between documents - similar topics cluster together.

Colors: Red = IMAGES, Teal = TEXT

View Visualization

Network Graph

Co-occurrence network of names in documents. Node size indicates mention frequency. Edges show which names appear together in the same documents.

Hover over nodes to see mention counts.

View Visualization

Document Distribution

Breakdown of documents by type (IMAGES vs TEXT) and volume. Shows the composition of the 69,290 document chunks.

TEXT-001 = 44.7% of all chunks (legal documents)

View Visualization

Resources

RAG Mining Report - Full findings with fact-checking and source citations
HuggingFace Dataset - Full embeddings (69,290 vectors, 768-dim)
Source Dataset - Original OCR'd documents
Visualization Summary - Statistics and methodology
Source Code - Python script used to generate visualizations

Epstein Files Visualizations

Dataset Statistics

Embedding Cluster Map

Network Graph

Document Distribution

Resources