2024 – Present
Team Members: Yu Fu (Georgia Tech / UCF), Denny Bromley, Vidya Setlur (Tableau Research)
DataWeaver is an interactive bidirectional authoring tool that connects text and visuals through seamless workflows for creating data stories.
Developed in collaboration with Tableau Research, the system bridges visualization-to-text and text-to-visualization generation using large language models (LLMs) such as ChatGPT.
If a user starts from visualizations, DataWeaver can help surface descriptive data facts (e.g., trends, comparisons, outliers) and draft accompanying narrative text.
If the user starts from text, the system suggests relevant charts and visual encodings aligned with the written content.
This closed authoring loop allows people to fluidly move between data and narrative, while maintaining direct editorial control.
The project highlights how human–AI collaboration can empower a wider range of people—not just data experts—to craft accurate, engaging, and trustworthy visual narratives.
News Coverage: Tableau Blog – Visualize, Narrate, Repeat: Closing the Loop on Data Storytelling.
Conference: Presented at EuroVis 2025, Luxembourg.