Examples play a critical role in both artificial intelligence and human creativity. Machine learning requires high-quality labeled examples to train models; in human creative tasks, examples serve as inspiration and reference points for generating new ideas. However, current investigations on data visualization design and generation have largely overlooked the role and nature of examples. AI models are often trained on visualization corpora with limited diversity, lacking fine-grained labels on the constituent parts; on the other hand, few studies examine why or how visualization creators incorporate examples into their workflows, and how examples influence visualization outcomes. Consequently, there exists no guidelines on building tools to support example-aided visualization design and implementation.

 

In this talk, I will discuss research projects from my group that seek to address these gaps: representations and models for interactive data visualizations, diverse visualization corpus with fine-grained semantic labels, empirical studies on the role of examples in visualization design, and interactive systems for visualization authoring. These projects aim to leverage diverse real-world visualization examples for enhancing both automation and human creativity in visualization design and generation.