Information Visualization with ChatGPT

Research output: Chapter in Book/Report/Conference proceedingsChapterpeer-review


We investigate the capabilities of the chatbot ``ChatGPT'' and the language model behind it for information visualization. For this purpose, we compare the creation of information visualizations with ChatGPT to other approaches for chart generation using natural language. We use ChatGPT with Jupyter Notebooks on one hand and a web application that uses the deep learning model OpenAI Codex to create visualizations from text input on the other. In addition, we use a prototype with a classical natural language processing (NLP) approach based on the Natural Language for Data Visualization (NL4DV) toolkit (with subtasks such as part-of-speech (POS) tagging, entity recognition, and dependency parsing). We investigate the advantages, disadvantages and suitability of the approach for the most common visualization types. The experiments with ChatGPT yield better results than with the other two variants with OpenAI Codex and NL4DV. Users can create more complex visualizations with the Deep Learning approaches but sometimes need help in finding the appropriate text input to solve the tasks.
Original languageEnglish
Title of host publicationArtificial Intelligence and Visualization: Advancing Visual Knowledge Discovery
EditorsBoris Kovalerchuk, Kawa Nazemi, Răzvan Andonie, Nuno Datia, Ebad Bannissi
Place of PublicationCham
Number of pages17
ISBN (Print)978-3-031-46549-9
Publication statusPublished - Apr 2024

Publication series

NameStudies in Computational Intelligence
ISSN (Print)1860-949X
ISSN (Electronic)1860-9503

Cite this