dto-btn / ssc-assistant

Second iteration of the SSC chatbot/assistant
MIT License
0 stars 0 forks source link

AAACT feedback (accessbility fixes) #131

Open GuillaumeTurcotte opened 3 weeks ago

GuillaumeTurcotte commented 3 weeks ago

THIS IS FEEDBACK RECEIVED BY AAACT TEAM BELOW

Users that require voice recognition in order to use their computer have an adaptive technology such as Dragon Naturally Speaking to do so. Therefore, when it comes to those users, it should be a priority to ensure that the SSC Assistant is compliant to the digital accessibility standards. We recommend using the EN 301 549 (2021), within this standard, the requirements that are relevant to this project is Chapter 9 Web which equates to the Web Content Accessibility Guidelines (WCAG) 2.1 level AA. Unfortunately, SSC Assistant currently does not seem to pass WCAG 2.1 level AA. For example, there are a couple of issues that were found: • Success Criterion 1.4.11: Non-text Contrast, the keyboard focus indicator must succeed a contrast of 3:1. Screenshot of the submit button: image

Success Criterion 1.3.1: Info and Relationships, aria-hidden is put onto a button that conveys important information. Aria-hidden is given to the button that submits the question to the assistant. This hides the semantic markup from adaptive technologies, causing for example, screen reader to read the button as simply “button”. HTML Code of the submit button:

Success Criterion 1.3.1: Info and Relationships, the relationship between the action buttons and the responses are not programmatically determined. For example, the screen reader reads when tabbing through the interface: o Copy button
o Good response button
o Bad response button
o Copy button
o Good response button
o Bad response button
o Copy button
o Regenerate button
o Good response button
o Bad response button
Screenshot of the action buttons:

image

Success Criterion 1.3.1: Info and Relationships, a heading level 6 is being used for the main heading of the page. We recommend starting with a heading level 1 instead. • There are a couple of user experience ameliorations that could be done as per the feedback of a screen reader that was shared with you (Luke-Robinett/ChatGPTAccessibilityEnhancementsChromeExtension: Adds screen reader accessibility enhancements to the ChatGPT user interface when using Google Chrome. (github.com)). For example: o Making the contents of the assistant’s responses be headings level 3 o Announce when GPT begins and finishes generating its response