Achieving web accessibility is essential to building inclusive digital experiences. However, accessibility issues are often identified only after a website has been fully developed, making them difficult to address. This paper introduces a Visual Studio Code plugin that integrates calls to a Large Language Model (LLM) to assist developers in identifying and resolving accessibility issues within the IDE, reducing accessibility defects that might otherwise reach the production environment. Our evaluation shows promising results: the plugin effectively generates functioning fixes for accessibility issues when the errors are correctly detected. However, detecting errors using a generic prompt-designed for broad applicability across various code structures-remains challenging and limited in accuracy.

翻译：实现网页无障碍对于构建包容性数字体验至关重要。然而，无障碍问题通常在网站完全开发后才被发现，导致修复困难。本文介绍了一款Visual Studio Code插件，该插件集成大型语言模型调用，以帮助开发者在集成开发环境中识别和解决无障碍问题，从而减少可能进入生产环境的无障碍缺陷。我们的评估显示出有希望的结果：当错误被正确检测时，该插件能有效生成针对无障碍问题的功能性修复方案。然而，使用为适应各种代码结构而设计的通用提示来检测错误仍然具有挑战性，且准确性有限。