A survey of visual insight mining: Connecting data and insights via visualization

Insight mining transforms complex data into actionable knowledge, enabling effective decision-making across diverse domains. Given the richness and interpretative power of visualizations, visual insight mining - the process of extracting meaningful insights from raw data through intuitive visual representations - has become increasingly vital. This survey systematically reviews the current landscape of visual insight mining, addressing the critical questions: “How can visualizations be generated from data?” and “How can insights be extracted from visualizations?” . Specifically, we delve into six distinct tasks (i.e., task decomposition, visualization generation, visualization recommendation, chart parsing, chart question answering, and insight generation) in the process of visual insight mining, and provide a comprehensive analysis of rule-based, learning-based, and large-model-based methods for each task. Based on the survey, we discuss current research challenges and outline future opportunities. By viewing visualization as a bridge in the data-to-insight path, this survey offers a structured foundation for further exploration in visual insight mining. • We provide a first systematic survey of visual insight mining, highlighting its role in turning complex data into actionable knowledge via the key stages Data2Vis and Vis2Insight. • We analyze six critical tasks (task decomposition, visualization generation, visualization recommendation, chart parsing, chart question answering, and insight generation) and categorize existing approaches into rule-based, learning-based, and large-model-based methods. • We argue that visualization serves as a critical bridge between raw data and insights, emphasizing its growing potential to fully close the data-insight gap in the multimodal era.