PlotThread: Creating Expressive Storyline Visualizations using Reinforcement Learning.

Area of Science:

• Computer Science
• Human-Computer Interaction
• Artificial Intelligence

Background:

• Storyline visualizations effectively present plot evolution and character interactions.
• Designing these visualizations is challenging due to the need to balance aesthetics and narrative constraints.
• Existing automated methods struggle with efficient design space exploration and flexible layout customization.

Purpose of the Study:

• To develop a reinforcement learning (RL) framework for AI-assisted storyline visualization design.
• To create PlotThread, an authoring tool for efficient exploration and customization of storyline layouts.
• To enable a mixed-initiative approach for collaborative design between AI agents and human users.

Main Methods:

• A reinforcement learning framework was developed to train an AI agent for storyline optimization.
• The PlotThread authoring tool was introduced, featuring flexible interactions for customization.
• A mixed-initiative design approach was employed, integrating AI and human designers on a shared canvas.

Main Results:

• The RL model demonstrated effectiveness in generating optimized storyline layouts.
• PlotThread facilitates efficient exploration of the storyline design space.
• Qualitative and quantitative experiments validated the RL model's performance.

Conclusions:

• The proposed RL framework and PlotThread tool significantly improve the efficiency and flexibility of storyline visualization design.
• Mixed-initiative design enhances the collaborative creation of complex narrative visualizations.
• This approach offers a powerful solution for users balancing aesthetic and narrative requirements in storyline visualization.