ExcelViZ: Automated Generation of High-Level, Adaptable Scatterplot Descriptions Based on a User Study

Digital charts enable quick and effective analysis of complex data, while they pose a barrier for people with visual impairments. Image descriptions are intended to provide equivalent access to the chart’s content, which is challenging and requires interpretations of the data, especially for complex data sets. Therefore, this paper investigates which key messages and categories sighted students perceive in scatterplots based on two user studies. First, we discussed five exemplary scatterplots with a group of sixteen students, who are familiar with digital accessibility and scatterplots in the academic context. We further analyzed this sample of scatterplots through an online survey with 222 participants, primarily students. We also compared the statistical correlations with the correlations reported by the participants, finding a high degree of agreement on outliers and trends. The results highlight the potential of image descriptions for accessible data analysis and form the basis for their automated generation.