Semantic Substrates
What is a Semantic Substrate?
A semantic substrate is a network visualization technique where node positions are determined by their attributes rather than network topology. The display is divided into meaningful regions based on node categories or properties, making it easy to see patterns within and between groups. Unlike force-directed layouts where position is arbitrary, every location in a semantic substrate conveys meaning.
Why Use Semantic Substrates?
Semantic substrates excel when you need to:
- Analyze by attributes: See how network structure relates to node properties (type, category, time, value).
- Compare groups: Understand connection patterns within and between different node categories.
- Reduce visual clutter: Eliminate edge crossings by strategic node placement within regions.
- Answer specific questions: Position nodes to directly address analytical questions about your data.
- Maintain mental models: Consistent, meaningful positions help users build understanding.
How It Works in GraphPolaris
Creating and exploring a semantic substrate in GraphPolaris is straightforward:
- Load your network data with node attributes (categories, types, values).
- Define regions based on one or two node attributes.
- Position nodes within regions using secondary attributes or spacing algorithms.
- Explore relationships: See edges connecting nodes within and across regions.
GraphPolaris supports flexible region definitions—categorical grids, temporal axes, or custom arrangements.
Visual Patterns
Understanding common visual patterns in semantic substrates helps you quickly interpret attribute-based network structure. Here are the key patterns to look for:
Dense Region (Active Category)
A region with many nodes indicates a category with high membership. Combined with many internal edges, it shows an active, interconnected group.
┌─────────────────┬─────────────────┐
│ Category A │ Category B │
│ │ │
│ ●──●──●──● │ ● │
│ │╲ │ ╱│ ╲│ │ │
│ ●──●──●──● │ ● ● │
│ │╱ │ ╲│ ╱│ │ │
│ ●──●──●──● │ │
│ │ │
└─────────────────┴─────────────────┘
Dense Sparse
What to look for: Regions filled with many nodes and internal connections. Compare density across regions to see category sizes.
Cross-Region Bridges
Edges spanning between regions show relationships across categories. Dense cross-region connections indicate strong inter-group relationships.
┌─────────────────┬─────────────────┐
│ Category A │ Category B │
│ │ │
│ ●──────────┼──────────● │
│ │ │ │ │
│ ●──────────┼──────────● │
│ │ │ │ │
│ ●──────────┼──────────● │
│ │ │
└─────────────────┴─────────────────┘
Cross-region edges
What to look for: Edges crossing region boundaries. Many parallel cross-region edges suggest systematic relationships between categories.
Isolated Region
A region with nodes but few or no edges (internal or external) indicates an isolated group—potentially disconnected from the main network.
┌─────────────────┬─────────────────┐
│ Connected │ Isolated │
│ │ │
│ ●──●──● │ ● ● │
│ ╲ │ ╱ │ │
│ ●──● │ ● ● ● │
│ │ │ │
│ ● │ ● │
│ │ │
└─────────────────┴─────────────────┘
What to look for: Regions with nodes but no connecting edges. May indicate data quality issues or genuinely separate groups.
Hub in Region
A node with many edges within its region (and possibly to other regions) is a local hub—a central entity within its category.
┌─────────────────┬─────────────────┐
│ Category A │ Category B │
│ │ │
│ ● ● │ │
│ ╲ ╱ │ ● │
│ ╲ ╱ │ ╱│╲ │
│ ●───◉───● │ ● ● ● │
│ ╱ ╲ │ │
│ ╱ ╲ │ │
│ ● ● │ │
└─────────────────┴─────────────────┘
Local hub
What to look for: A node with many radiating edges within its region. Compare hubs across regions to find category leaders.
Bipartite Structure
When edges only connect nodes in different regions (not within), it reveals a bipartite or multi-partite structure.
┌─────────────────┬─────────────────┐
│ Type A │ Type B │
│ │ │
│ ●──────────┼──────────● │
│ │ ╱ │
│ ●──────────┼────────● │
│ │ ╱ │
│ ●──────────┼──────● │
│ │ │
│ (no internal │ (no internal │
│ edges) │ edges) │
└─────────────────┴─────────────────┘
What to look for: Edges only between regions, never within. Common in user-item, author-paper, or other two-mode networks.
Temporal Progression
When regions represent time periods, patterns across columns reveal temporal evolution of network structure.
┌──────────┬──────────┬──────────┬──────────┐
│ 2020 │ 2021 │ 2022 │ 2023 │
│ │ │ │ │
│ ● │ ●─────┼────● │ ● │
│ │ ╱│ │ │╲ │ ╱│╲ │
│ ●─────┼──● │ │ │ ●───┼──● │ ● │
│ │ │ │ │ │ │ │
│ │ ●─────┼────●─────┼────● │
│ │ │ │ │
└──────────┴──────────┴──────────┴──────────┘
Edges show evolution over time
What to look for: How nodes and edges appear, persist, or disappear across time-based regions.
Clustered Sub-regions
Within a region, natural clusters of tightly connected nodes may emerge, revealing sub-structure within categories.
┌─────────────────────────────────────┐
│ Category A │
│ │
│ ┌─────────┐ ┌─────────┐ │
│ │ ●──●──● │ │ ●──● │ │
│ │ ╲ │ ╱ │ │ │╲ │ │ │
│ │ ●──● │ │ ●──● │ │
│ └─────────┘ └─────────┘ │
│ Cluster 1 Cluster 2 │
│ │
└─────────────────────────────────────┘
What to look for: Groups of tightly connected nodes within regions. May indicate sub-categories or communities.
Attribute Gradient
When one axis represents a continuous attribute (value, time, rank), node positions along that axis encode quantitative information.
│ High ● ●
│ │╲ ╱
│ Value ● ╲ ╱ ●
│ ╲ ╲ ╱ ╱
│ ╲ ● ╱
│ ╲│╱
│ Low ●────●────●
└──────────────────────────
Category →
What to look for: How nodes distribute along quantitative axes. Edges connecting across value ranges show relationships between different attribute levels.
Semantic substrates provide meaningful, attribute-driven network views—and with GraphPolaris, you can flexibly define regions to answer specific analytical questions about your connected data.