Automatic summary generation by data extraction of sunburst charts

Abstract

Sunburst charts are useful for visualizing hierarchical data structures. However, they can be challenging to interpret because of their nested hierarchy and circular layout. Viewers must determine the color and size and of each segment and navigate through various hierarchy levels, which can be confusing, especially with many or unevenly distributed categories. To address these challenges, this research proposes a method to extract summary data from sunburst charts, enabling quick identification of key trends and patterns. This approach is particularly beneficial for business analysts and marketers who need to analyze large datasets efficiently. By automating the extraction process, analysts can focus their efforts on areas of significance, saving time and improving efficiency. Existing literature indicates a gap in tools for extracting data and generating summaries of sunburst charts. To fill this gap, the author implemented a system that leverages Machine Learning (ViT models), image processing, optical character recognition (OCR), and LLM. While several methods could have been used for this project, the chosen technologies offer a robust solution for data extraction and summary generation from sunburst charts