Atlas: Grammar-based Procedural Generation of Data Visualizations

Zhicheng Liu, Chen Chen, Francisco Morales, Yishan Zhao

Abstract: We present Atlas, a procedural grammar for constructing data visualizations. Unlike most visualization grammars which use declarative specifications to describe visualization components, Atlas exposes the generative process of a visualization through a set of concatenated high-level production rules. Each of these rules describes how an input graphical object is created, transformed, or joined with abstract data to derive an output object. The visualization state can thus be inspected throughout the generative process. We demonstrate Atlas’ expressivity through a catalog of visualization designs, and discuss the trade-offs in its design by comparing it to state-of-the-art grammars.