KB4VA: A Knowledge Base of Visualization Designs for Visual Analytics

Visual analytics (VA) systems have been widely used to facilitate decision-making and analytical reasoning in various application domains. VA involves visual designs, interaction designs, and data mining, which is a systematic and complex paradigm. In this work, we focus on the design of effective visualizations for complex data and analytical tasks, which is a critical step in designing a VA system. This step is challenging because it requires extensive knowledge about domain problems and visualization to design effective encodings. Existing visualization designs published in top venues are valuable resources to inspire designs for problems with similar data structures and tasks. However, those designs are hard to understand, parse, and retrieve due to the lack of specifications. To address this problem, we build KB4VA, a knowledge base of visualization designs in VA systems with comprehensive labels about their analytical tasks and visual encodings. Our labeling scheme is inspired by a workshop study with 12 VA researchers to learn user requirements in understanding and retrieving professional visualization designs in VA systems. The theme extends Vega-Lite specifications for describing advanced and composited visualization designs in a declarative manner, thus facilitating human understanding and automatic indexing. To demonstrate the usefulness of our knowledge base, we present a user study about design inspirations for VA tasks. In summary, our work opens new perspectives for enhancing the accessibility and reusability of professional visualization designs.