Chart Accessibility: A Review of Current Alt Text Generation

Abstract

Charts are valuable visual representations of data, provide an easy overview, and reveal important aspects of the data, such as trends and outliers, which are usually difficult to notice in textual representations. Charts are important for sighted readers but are inaccessible to screen reader users without the help of alternative text (alt text). Thus far, alt text has to be written manually, but recent advancements in deep learning models can support this process by relying on automatic chart understanding models. While numerous studies have explored automatic chart understanding tasks, there remains a lack of comprehensive reviews specifically focused on automatic alt text generation for screen reader users. In this study, we collected 38 papers and reviewed them according to the alt text generation workflow: target alt text content, methods for automatic alt text generation, and available applications that can assist alt text generation. Furthermore, we discuss the remaining challenges and outline the directions for future research in this field. This study is the first comprehensive survey that specifically focuses on the generation of alt text for impaired individuals. It serves as a resourceful overview for researchers and practitioners working on alt text generation, emphasizing the need for the continued development of chart accessibility.