Assesment_for_AI_ALLY_Report - RutgersGRID/VIDAHub GitHub Wiki

ALLY Accessibility Report

Table of Contents

This is an attempt to breaddown and better understand the ALLY Accessibility Report.

The Ally Tool looks for a large list of potential issues.

It can review files in these formats:

  • PDF files
  • Microsoft® Word files
  • Microsoft® PowerPoint® files
  • OpenOffice/LibreOffice files
  • Uploaded HTML files
  • Image files (JPG, JPEG, GIF, PNG, BMP, TIFF)
  • WYSIWYG/VTBE content
  • YouTubeTM videos embedded in WYSIWYG/VTBE content

Accessibility Review

Report Overview

The top part of the page shows the overall score and a breakdown of all the types of content in the course

Quick Links

  • On the right, the report also has links to two options to help you quickly make improvements to the site.
    • First is a link to easy-to-fix content (usually missing image descriptions). Just click the Start button to jump to a list of all content ALLY considered "easy-to-fix".
    • Second is a link to the lowest scoring (or least accessible) content. The items here will be the least accessible to end users, and fixing these issues will likely provide the best "return on investment" in improving your site's accesibility score, and therefore, how accessible your site will be to those who face accessibility issues.

Issues List

Lower on the main page, issues are listed, sorted by frequency/amount of content with that issue (e.g., contrast issues, missing alt text, etc.).

  • Each row is a type of issue and how often that issue occurs in the canvas site.

  • The list defaults to sorting bt issue severity and then most common issues, but you can click on column headers to change the sort.

  • Click on a row to go to a list of the content with that specific issue.

Reviewing an issue category

  • After clicking on an issue category, you will be presented with a list of all the content that has that issue. Here is a list of images that are missing descriptions.

  • Clicking on an item/row will show you additional details about that piece of content.

Reviewing Low Scoring Content

Clicking the low scoring quick link at the top right of the main page of the report takes you to a list of the least accessibly content in the site.

  • Clicking on an item/row will show you additional details about that piece of content.

Reviewing Easiest-to-Fix Content

  • Clicking on an item/row will show you additional details about that piece of content.

Content in a specific issue category

  • After clicking to view a specific piece of content, you'll be shown the content, its accessibility score, and the issue.
  • There will be buttons on the upper right that will give you additional information.
  • In this example, the issue is images that are missing descriptions or alt tags.
  • In some cases, like this example, you may be able to fix the issue immediate, such as for missing descriptions for images. You can enter a description here to update this piece of content.
  • For easier fixes, like adding a description, you can enter the description right here.
    • Note: It is suggested you read through the suggestions on how to write a good description before attempting to do so.
  • In other cases, such as PDFs or other documents with other problems, ALLY will attempt to explain the issue and why it's important, and offer suggestions on making improvements.
Getting more information

Buttons lead to additional information about the issues for each piece of content. Here are examples from an untagged PDF.

  • Click "What this means" to learn more.

  • Click "How to tag a PDF" to learn more about the content you're viewing. Below is the information provided for an untagged PDF.

  • NOTE: If you do not have the original version of the PDF (to edit) and you do not have a link to the original file, ALLY will tell you you cannot create a tagged PDF.

Here is another example, for an infographic that's missing a description.

Example Issues
  • Scanned but not OCRed
  • Documents with contrast issues

If there are more than one issue, there will be a navigation tool at the top of the screen. Clicking the "eye" icon will turn off the highlight of the issues.

  • Links often show up as contrast issues.
Another example (with course name removed)
Missing Image Description button results
How to write good image descriptions

Using AI to correct for accessibility issues

Each type of issue may require a unique solution. The tool to add image descriptions won't work for missing headers in a word document, for example.

Creating (Good) Image Descriptions with AI

Automation will require a delicate touch. For example, one of the easier things to fix is missing image descriptions, but step 3 in the ALLY provided "How to write a good description" shows why that's harder than just image recognition.

Here's the text.

Keep it brief and context-specific Consider the purpose of the image. Image descriptions can vastly differ depending on the context and meaning it tries to convey. Keep your descriptions concised and relevant.

In history class A field covered with poppies. Poppies have become a symbol of remembrance for World War I.

In art class A painting by Van Gogh in Post-Impressionism using dramatic brush strokes. It shows a field with poppies in the foreground, trees on the background, and dark skies.

This means that if we were to build a tool, it would also need some context about the course. We still may be able to automate this to some degree, but we'll need to add a form of some sort that asks the site creator/site editor for context to apply and we'll need to carefully build a prompt that is contextually aware of the image.

Even with that additional context, other images might be far more complex to automate a description, like this instruction for loading a printer:

The actual instruction here is "Flip the feeder guard forward." but if that's in the text and then we also have it in the alt text, it'll be read twice by screen readers.