BOLT: A Natural Language Interface for Dashboard Authoring

Abstract

Authoring dashboards is often a complex process, requiring expertise in both data analysis and visualization design. With current tools, authors lack the means to express their objectives for creating a dashboard (e.g., summarizing data changes or comparing data categories), making it difficult to discover and assemble content relevant to the dashboard. Addressing this challenge, we propose the idea of employing natural language (NL) for dashboard authoring with a prototype interface, BOLT. In this paper, we detail BOLT's design and implementation, describing how the system maps NL utterances to prevalent dashboard objectives and generates dashboard recommendations. Utilizing BOLT as a design probe, we validate the proposed idea of NL-based dashboard authoring through a preliminary user study. Based on the study feedback, we highlight promising application scenarios and future directions to support richer dashboard authoring workflows.