BiSet: Semantic Edge Bundling with Biclusters for Sensemaking

Area of Science:

• Data Analytics
• Information Visualization
• Human-Computer Interaction

Background:

• Identifying coordinated relationships is crucial in data analytics.
• Current methods for detecting these relationships are manual and require significant cognitive effort.
• Visualizing individual relationships often leads to cluttered displays, hindering the discovery of coordinated patterns.

Purpose of the Study:

• To introduce BiSet, a novel visual analytics technique for interactive exploration of coordinated relationships.
• To address the limitations of existing methods in identifying complex, coordinated patterns.
• To enable analysts to infer semantic insights about coordinated activities more effectively.

Main Methods:

• Modeling coordinated relationships as biclusters.
• Algorithmically mining biclusters from datasets.
• Visualizing biclusters as bundled edges between sets of related entities.
• Introducing a new "in-between" layer to organize bundle objects and entities.

Main Results:

• BiSet effectively visualizes coordinated relationships by representing them as bundles.
• Bundles organize entities and reveal their membership, facilitating sensemaking.
• Interactive manipulation of bundles allows users to explore and organize related entities.
• A usage scenario demonstrates BiSet's utility in text analytics for coordinated relationship exploration.

Conclusions:

• BiSet provides an effective visual analytics technique for discovering coordinated relationships.
• The bundled edge visualization and interactive features enhance the exploration of complex data patterns.
• BiSet supports task-oriented semantic insights and improves sensemaking in data analytics, particularly in text analysis.