DynaVis: Dynamically Synthesized UI Widgets for Visualization Editing

Priyan Vaithilingam,Elena L. Glassman,Jeevana Priya Inala,Chenglong Wang

2024-01-20

Abstract:Users often rely on GUIs to edit and interact with visualizations - a daunting task due to the large space of editing options. As a result, users are either overwhelmed by a complex UI or constrained by a custom UI with a tailored, fixed subset of options with limited editing flexibility. Natural Language Interfaces (NLIs) are emerging as a feasible alternative for users to specify edits. However, NLIs forgo the advantages of traditional GUI: the ability to explore and repeat edits and see instant visual feedback.

Human-Computer Interaction

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that while existing visual editing tools provide rich editing options, they also bring the trouble of complex interfaces to users, or simplify the interfaces to reduce complexity but sacrifice editing flexibility. Specifically, the author points out:

1. **Challenges of complex interfaces**: Modern interactive visual creation tools (such as Tableau, PowerBI, etc.) have greatly reduced the effort of creating initial visualizations from data. However, the "intelligent default values" provided by these tools are often sufficient for exploratory analysis. When users need to make fine - tuned adjustments to the visualizations to better convey insights or meet specific analysis goals, these default settings become insufficient. Users need to distinguish which options can achieve the desired editing effects, which requires expertise in low - level visualization grammar. In addition, finding these editing options may require searching through the multi - layer menus and panels of the tool, which is a challenge for users without sufficient tool expertise.

2. **Limitations of natural language interfaces (NLI)**: Although natural language interfaces (NLI) allow users to describe editing effects through natural language, thus solving the discovery and navigation challenges, they give up the advantages of traditional graphical user interfaces (GUI), especially the ability to perform fine - grained editing, obtain immediate visual feedback, and quickly undo and re - apply edits.

To address the above challenges, the author designed a new interaction method - dynamic widgets, and implemented this method by developing a tool named DynaVis. DynaVis combines natural language interfaces with interactive GUI widgets, enabling users to describe editing tasks through natural language. DynaVis will then perform the edits and generate a persistent widget for users to further modify. This method aims to reduce the execution gap (i.e., the difference between user intentions and system capabilities), while enhancing interactivity, allowing users to enjoy the convenience brought by natural language interfaces and also experience the immediate feedback and editing flexibility of traditional GUIs.