Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Macro-Queries: An Exploration into Guided Chart Generation from High Level Prompts

Christopher J. Lee, Giorgio Tran, Roderick Tabalba, Jason Leigh, Ryan Longman

TL;DR

The paper tackles making data visualization accessible to novices by enabling chart generation from high-level prompts called macro-queries. It presents a modular, LLM-driven pipeline that maps macro-queries to visualizations through prompting guided by Abela's Chart Taxonomy, a text-to-SQL transformation layer, and chart templates, augmented by retrieval-augmented generation. The authors define macro-queries, propose an end-to-end pipeline, and demonstrate preliminary feasibility across diverse chart types, while discussing design rationales, limitations, and potential improvements. The work suggests that macro-queries can broaden access to data exploration across domains, though further work is needed to improve robustness, chart coverage, and interaction fidelity.

Abstract

This paper explores the intersection of data visualization and Large Language Models (LLMs). Driven by the need to make a broader range of data visualization types accessible for novice users, we present a guided LLM-based pipeline designed to transform data, guided by high-level user questions (referred to as macro-queries), into a diverse set of useful visualizations. This approach leverages various prompting techniques, fine-tuning inspired by Abela's Chart Taxonomy, and integrated SQL tool usage.

Macro-Queries: An Exploration into Guided Chart Generation from High Level Prompts

TL;DR

The paper tackles making data visualization accessible to novices by enabling chart generation from high-level prompts called macro-queries. It presents a modular, LLM-driven pipeline that maps macro-queries to visualizations through prompting guided by Abela's Chart Taxonomy, a text-to-SQL transformation layer, and chart templates, augmented by retrieval-augmented generation. The authors define macro-queries, propose an end-to-end pipeline, and demonstrate preliminary feasibility across diverse chart types, while discussing design rationales, limitations, and potential improvements. The work suggests that macro-queries can broaden access to data exploration across domains, though further work is needed to improve robustness, chart coverage, and interaction fidelity.

Abstract

This paper explores the intersection of data visualization and Large Language Models (LLMs). Driven by the need to make a broader range of data visualization types accessible for novice users, we present a guided LLM-based pipeline designed to transform data, guided by high-level user questions (referred to as macro-queries), into a diverse set of useful visualizations. This approach leverages various prompting techniques, fine-tuning inspired by Abela's Chart Taxonomy, and integrated SQL tool usage.
Paper Structure (39 sections, 5 figures, 9 tables)

This paper contains 39 sections, 5 figures, 9 tables.

Table of Contents

  1. Introduction and Background
  2. Macro-Queries
  3. LLM Approach
  4. Contributions
  5. Related Work
  6. Architectural Approach
  7. Early Attempts
  8. Current Architecture
  9. Walk-through
  10. Modules
  11. API and Webapp:
  12. (Optional) Reiteration (A in Figure \ref{['fig:model']}):
  13. CSV Decomposition/ Analysis (B in Figure \ref{['fig:model']}):
  14. (Optional) Attribute Filter (C in Figure \ref{['fig:model']}):
  15. (SQL) Transformation (D in Figure \ref{['fig:model']}):
  16. ...and 24 more sections

Figures (5)

  • Figure 1: Andrew Abela's Chart Taxonomyabela with chart templates superimposed and alteration from components of components to treemap chart since both charts embody equivalent principles.
  • Figure 2: Model Architecture, where the inputs are a CSV and user prompt and the output is a JSON describing how to construct the visualization with the relevant CSV.
  • Figure 3: Home page of the web app utilizing the API
  • Figure 4: Response with reasoning at each step if applicable invoked by the macro-query: "What things should I sell?"
  • Figure 5: Continuation of Figure \ref{['fig:demosplit1']}'s reasoning with a interactive variable width column chart generated based on Abela's Chart Taxonomy