GistVis is a proof-of-concept research prototype that aims to provide in-situ augmentation of data-rich documents. GistVis blends the power of Large Language Model (LLM) and Visualization Design Knowledge to automatically generate word-scale visualizations (WSV) within documents.
GistVis is also actively evolving as the technical toolchain evolves (prototype implementation in releases/gistvis-prototype branch). Building on the research prototype's modular design, our team's immediate next step is separating and improving the WSV module into a React component library to facilitate research and application with WSVs. We welcome constructive criticisms and contributions.
We are working on publisishing a website for our project. Stay tuned!
If you find GistVis useful, please cite our paper:
@article{zou2025gistvis,
title = {GistVis: Automatic Generation of Word-scale Visualizations from Data-rich Documents},
author = {Zou, Ruishi and Tang, Yinqi and Chen, Jingzhu and Lu, Siyu and Lu, Yan and Yang, Yingfan and Ye, Chen},
journal = {arXiv preprint arXiv:2502.03784},
year = {2025}
}This work is currently conditionally accepted to ACM CHI 2025.