ACCESS: Prompt Engineering for Automated Web Accessibility Violation Corrections

TL;DR

This work targets the pervasive problem of WCAG violations by introducing ACCESS, a pipeline that automatically corrects web accessibility violations through real-time DOM modifications guided by foundation models. It constructs the ACCESS Benchmark to quantitatively evaluate corrections, using a five-level severity scale and aggregate scores, and explores three prompt-engineering strategies (ReAct, Few-Shot Guided, and Transformer-based prompts) to generate corrected HTML that is applied to the DOM. The results show a substantial reduction in accessibility violations, with a best-case reduction of about 51% in severity using ReAct prompting on GPT-3.5-turbo-16K, and a GPT-4 baseline offering additional gains on a smaller subset. The work highlights practical implications for end-user accessibility tools and outlines future directions, including dataset expansion and multimodal approaches, to further automate and generalize web accessibility repairs.

Abstract

With the increasing need for inclusive and user-friendly technology, web accessibility is crucial to ensuring equal access to online content for individuals with disabilities, including visual, auditory, cognitive, or motor impairments. Despite the existence of accessibility guidelines and standards such as Web Content Accessibility Guidelines (WCAG) and the Web Accessibility Initiative (W3C), over 90% of websites still fail to meet the necessary accessibility requirements. For web users with disabilities, there exists a need for a tool to automatically fix web page accessibility errors. While research has demonstrated methods to find and target accessibility errors, no research has focused on effectively correcting such violations. This paper presents a novel approach to correcting accessibility violations on the web by modifying the document object model (DOM) in real time with foundation models. Leveraging accessibility error information, large language models (LLMs), and prompt engineering techniques, we achieved greater than a 51% reduction in accessibility violation errors after corrections on our novel benchmark: ACCESS. Our work demonstrates a valuable approach toward the direction of inclusive web content, and provides directions for future research to explore advanced methods to automate web accessibility.

Figures (6)

• Figure 1: The big picture approach of our research. By automating accessibility violation corrections, we aim to eliminate the inefficiency and inaccuracy associated with manual error correction.
• Figure 2: Three example visual website violations that could affect web accessibility for individuals with impairments. (a) Contrast between foreground and background colors do not meet WCAG 2AA minimum contrast ratio thresholds. (b) Button does not have discernable text. (c) All page content is not contained by landmarks.
• Figure 3: Overview of the violation correction approach. (a) The violations dataset includes columns extracted from Playwright accessibility tests, used for prompt generation. (b) We generate the prompt using prompt engineering techniques and traverse the dataset to obtain corrected HTML tags from LLMs. (c) Our benchmark compares initial and final severity scores of errors in the DOM.
• Figure 4: Example ReAct prompt messages for a color contrast error. Additional example errors were included in the system message for other prompting techniques.
• Figure 5: The DOM correction process using prompt engineering and LLMs.