Graphiti: Interactive Specification of Attribute-Based Edges for Network Modeling and Visualization.

Area of Science:

• Computer Science
• Data Visualization
• Human-Computer Interaction

Background:

• Network visualizations are crucial for understanding complex relationships and identifying patterns in data.
• Existing tools often require programming expertise or complex queries for network creation from raw data.
• Current interaction techniques for network modeling assume prior knowledge of data attributes for defining relationships.

Purpose of the Study:

• To propose and evaluate a new interaction technique for network modeling that lowers the barrier to entry for users.
• To enable users to create network visualizations by demonstrating desired network structures rather than specifying explicit rules.
• To facilitate the creation and updating of network models from tabular data.

Main Methods:

• Developed a prototype tool, Graphiti, implementing a demonstration-based interaction technique.
• Users select example nodes and links to indicate desired network structures.
• The system infers and recommends conditions for generating the network based on user demonstrations.

Main Results:

• Graphiti successfully models networks from tabular datasets using the proposed interaction technique.
• The demonstration-based approach simplifies network creation, especially when data attributes are unknown or numerous.
• The tool effectively supports updating existing networks with new edge types.

Conclusions:

• The demonstration-based interaction technique offers an intuitive and efficient method for network modeling.
• This approach democratizes network visualization by reducing reliance on technical expertise.
• Graphiti demonstrates the practical application and benefits of this novel interaction paradigm.