Leveraging Foundation Models for Crafting Narrative Visualization: A Survey

Area of Science:

• Computer Science
• Data Visualization
• Artificial Intelligence

Background:

• Narrative visualization makes complex data accessible through storytelling.
• Foundation models offer advanced capabilities like NLP and content generation for data analysis.
• Integrating foundation models into narrative visualization is an emerging research area.

Purpose of the Study:

• To survey and analyze the application of foundation models in narrative visualization.
• To propose a reference model for understanding the role of foundation models in creating visual narratives.
• To identify key tasks and challenges in this interdisciplinary field.

Main Methods:

• Systematic literature review of 66 relevant papers.
• Development of a four-phase reference model: Analysis, Narration, Visualization, and Interaction.
• Identification and categorization of eight specific tasks where foundation models are applied.

Main Results:

• Foundation models can be progressively integrated across the narrative visualization pipeline.
• Eight key tasks, including Insight Extraction and Authoring, are detailed with associated foundation model applications.
• A structured overview of strengths, weaknesses, and research opportunities is provided for each task.

Conclusions:

• Foundation models significantly enhance the creation of narrative visualizations.
• The proposed model provides a framework for understanding and advancing research in AI-driven visual storytelling.
• This work guides researchers and practitioners in leveraging foundation models for impactful data communication.