Analysis & Interpretation

Set-Level Visualizations

Document set views aggregate similarity data across multiple witnesses. These interactive visualizations help identify document relationships.

All visualisations rely on the same underlying data: each document holds a number of extracted image regions, and each region is linked to a number of similar regions in other documents. This link is stored as a pair of images, associated with a similarity score and a category.

Filtering Options

All visualizations support filtering using sidebar controls:

• Document: Filter by document to focus on specific witnesses
• Category: Filter by similarity category (exact, partial, semantic)
• Score: Adjust score thresholds to show only high-confidence matches

Clusters

Document set clusters group images that are linked by at least on pair of similarities whose score exceeds the selected threshold.

Matrix Visualization

The similarity matrix shows pairwise document comparisons. Color intensity indicates the number of similar regions between documents.

Click on a cell to open a detailed matrix for that document pair. Axes show ordered illustrations; intensities reflect similarity ranks or annotations.

This view reveals whether copied illustrations maintain their original order, appear clustered in specific sections, or follow reorganization patterns.

Document Network Graph

Documents appear as nodes in an interactive graph. Edge width corresponds to the number of shared regions. Node size reflects connectivity.

Select nodes to display a table of all connected regions. This makes it easy to identify visual shifts, such as systematic mirroring.

Region Network Graph

At the region level, the network displays all extracted regions from a document set. Nodes represent individual regions; edges represent correspondences.

• Node size: proportional to the number of connections
• Edge width: reflects similarity score (when available)
• Selection: click nodes to view images and metadata

Large nodes indicate commonly copied images. Isolated nodes represent unique illustrations.

Stemma editor

The stemma editor allows users to try different ways to connect document within the set and test their hypothesis by looking at regions of interest of the documents through several visualization tools.