Problem
Research questions and friction points this paper is trying to address.
data visualization
intermediate language
semantic specification
visualization configuration
brittle defaults
Flint: A Semantics-Driven Data Visualization Intermediate Language
🤖 AI Summary
This work proposes Flint, a semantic-driven intermediate language that overcomes the limitations of traditional visualization systems, which often rely on manual configuration or shallow data inference, resulting in fragile and verbose specifications. Flint introduces a hierarchical data semantic model to structurally capture field semantics, enabling automatic generation of robust, portable, and high-quality chart configurations without requiring users to specify low-level details. By integrating semantic modeling, configuration compilation optimization, and multi-target syntax translation, Flint supports mainstream visualization backends such as Vega-Lite, Apache ECharts, and Chart.js. The system significantly streamlines the authoring workflow while preserving visual fidelity and ensuring cross-library compatibility, effectively meeting the needs of both human users and AI agents.
📝 Abstract
We present Flint, an intermediate language that enables authors to create high-quality visualizations from concise, semantics-driven specifications without explicitly configuring low-level parameters such as scales, axes, and formatting. Unlike prior systems that infer default configurations from surface-level data representations, often producing brittle choices, Flint introduces a hierarchical data semantic model that allows users to specify the meanings of data fields structurally and helps the compiler derive appropriate visualization configurations. From a concise specification, the system generates and optimizes library-agnostic visualization configurations and translates them into complete, executable specifications for multiple target grammars, including Vega-Lite, Apache ECharts, and Chart.js. We demonstrate that Flint simplifies the authoring process without compromising on visual quality, and it is an effective intermediate language for both humans and AI agents to create visualizations.
Innovation
Methods, ideas, or system contributions that make the work stand out.
semantics-driven
intermediate language
hierarchical data semantic model
visualization compilation
library-agnostic
🔎 Similar Papers
No similar papers found.
