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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zou, Ruishi, Tang, Yinqi, Chen, Jingzhu, Lu, Siyu, Lu, Yan, Yang, Yingfan, Ye, Chen
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910816750534656
author Zou, Ruishi
Tang, Yinqi
Chen, Jingzhu
Lu, Siyu
Lu, Yan
Yang, Yingfan
Ye, Chen
author_facet Zou, Ruishi
Tang, Yinqi
Chen, Jingzhu
Lu, Siyu
Lu, Yan
Yang, Yingfan
Ye, Chen
contents 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).
format Preprint
id arxiv_https___arxiv_org_abs_2502_03784
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle GistVis: Automatic Generation of Word-scale Visualizations from Data-rich Documents
Zou, Ruishi
Tang, Yinqi
Chen, Jingzhu
Lu, Siyu
Lu, Yan
Yang, Yingfan
Ye, Chen
Human-Computer Interaction
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).
title GistVis: Automatic Generation of Word-scale Visualizations from Data-rich Documents
topic Human-Computer Interaction
url https://arxiv.org/abs/2502.03784