Document Network

40466 documents positioned by embedding similarity — documents close together share similar content

Color by:
-3-3-2-2-1-100112233
40466 documents plotted by semantic similarity. Documents are grouped into 8 clusters.

About this visualization

Each point represents a document from the corpus. Documents are positioned using Principal Component Analysis (PCA) on 768-dimension embeddings generated by Google EmbeddingGemma. Documents with similar semantic content appear closer together.

Clusters are computed via K-Means (k=8) on the full embedding space. Cluster boundaries in 2D are approximate — the true cluster separation exists in the original 768 dimensions.

  • Color by Cluster — Groups documents by semantic similarity. Hover over points to see document IDs.
  • Color by Source — Shows documents colored by their origin dataset (FOIA, DOJ, etc.).