Falx: Synthesis-Powered Visualization Authoring

Abstract

Modern visualization tools aim to allow data analysts to easily create exploratory visualizations. When the input data layout conforms to the visualization design, users can easily specify visualizations by mapping data columns to visual channels of the design. However, when there is a mismatch between data layout and the design, users need to spend signifcant efort on data transformation.We propose Falx, a synthesis-powered visualization tool that allows users to specify visualizations in a similarly simple way but without needing to worry about data layout. In Falx, users specify visualizations using examples of how concrete values in the input are mapped to visual channels, and Falx automatically infers the visualization specifcation and transforms the data to match the design. In a study with 33 data analysts on four visualization tasks involving data transformation, we found that users can effectively adopt Falx to create visualizations they otherwise cannot implement.

Untidy Data: The Unreasonable Effectiveness of Tables

“…For example, permitting the native wide format, preserving both explicit and implicit associations from spatial structure such as aligned columns, linked tables and summary views, or including user-generated markup and marginalia. An example of affording flexibility in data structures is the Falx system [86], which attempts to infer ad hoc data transformations in order to generate custom visualizations rather than working around an assumed single and immutable data structure.…”

WebRobot: Web Robotic Process Automation using Interactive Programming-by-Demonstration

“…WebRobot adopts a human-in-theloop interaction model which has shown to be an effective way to incorporate human inputs when training AI systems in the HCI community [19,45]. This model has been used in the context of program synthesis, mostly interactive PBE [23,41,43,51,60,65]. In contrast, our work incorporates human inputs in PBD and proposes a new human-in-the-loop model.…”

Synthesizing Analytical SQL Queries from Computation Demonstration

“…Sickle is implemented in Python, the user demonstration is provided via spreadsheet; however, with some additional engineering effort, Sickle can work with a drag-and-drop interface [41] (as illustrated in Section 2) for smoother experience. In the synthesis process, Sickle enumerates join predicates (used in join and left_join) based on primary and foreign keys of the table to avoid generating unnatural predicates like "T1 Join T2 On T1.id < T2.age"; for the same reason, Sickle does not invent new constants for filter and considers only those provided by the user.…”

“…Providing an example requires users to manually find all the values that belong to a specific group and compute the aggregated value from them. This is costly when groups are created by partitioning and often results in erroneous examples (as reported in [41]). • Limited synthesis efficiency: Because examples capture only output values, the synthesizer needs to "guess" the functions used in the computation.…”