GistVis: Automatic Generation of Word-scale Visualizations from Data-rich Documents

TL;DR

GistVis is an automated pipeline that extracts and visualizes data insights from text in data-rich documents, improving user understanding and reducing mental effort through word-scale visualizations.

Contribution

It introduces GistVis, a novel system combining large language models and visualization design to generate word-scale visualizations from textual data descriptions.

Findings

GistVis generates satisfactory visualizations as per user feedback.

User study shows +5.6% accuracy in understanding data.

GistVis reduces mental demand and perceived effort significantly.

Abstract

Data-rich documents are ubiquitous in various applications, yet they often rely solely on textual descriptions to convey data insights. Prior research primarily focused on providing visualization-centric augmentation to data-rich documents. However, few have explored using automatically generated word-scale visualizations to enhance the document-centric reading process. As an exploratory step, we propose GistVis, an automatic pipeline that extracts and visualizes data insight from text descriptions. GistVis decomposes the generation process into four modules: Discoverer, Annotator, Extractor, and Visualizer, with the first three modules utilizing the capabilities of large language models and the fourth using visualization design knowledge. Technical evaluation including a comparative study on Discoverer and an ablation study on Annotator reveals decent performance of GistVis. Meanwhile, the user study (N=12) showed that GistVis could generate satisfactory word-scale visualizations, indicating its effectiveness in facilitating users' understanding of data-rich documents (+5.6% accuracy) while significantly reducing their mental demand (p=0.016) and perceived effort (p=0.033).