How Artificial Intelligence is Improving Digital Accessibility

Published October 28, 2022

Artificial intelligence is an evolving technology with great potential to enhance digital accessibility. More persons with disability can now access services in the digital world thanks to AI and machine learning. Let’s look at how AI is improving digital accessibility today.

Language caption and translation

Automated Speech Recognition (ASR) algorithms help provide subtitles and captions for video content, making it more accessible to those who need an alternative to audio. This technology also helps with self-learning, leading to improved captions over time. Microsoft Translator is one such dynamic technology, which is AI-enabled and helps provide captioning solutions. 

Another major technology is Google Translate, which uses a system called Google Neural Machine Translation (GNMT). This helps improve the real-time accuracy of its translation, reducing errors by as much as 55 to 85%. It provides more effective, nuanced idea- and sentence-based translation as opposed to word-based translation.

Image recognition

Individuals with visual impairment often face barriers to accessing images on the web or digital devices. Image processing powered by AI can be a valuable tool in these cases. 

Google Vision API is an advanced algorithm that utilizes neural networks for image recognition. Users with vision impairment can use this tool to identify images and determine if an image is labeled as safe. This technology is still in its nascent stages, but there are high hopes for its future application in image recognition.

Facial recognition

Logging into a digital device usually requires authentication, such as providing the correct PIN or password. However, cutting-edge artificial intelligence technologies are making it possible to use facial recognition instead. Facial recognition technology improves the accessibility of security features for those who may face barriers with input-based methods.

Many websites require a CAPTCHA to verify that a human is trying to access the site. But CAPTCHA can be inaccessible to many users, so some developers are experimenting with facial recognition technology as a replacement

Summarizing information

The internet has plenty of video and image content, but text content is still the most common. Screen readers serve as an effective tool for accessing text content, but long-form text can still be challenging to access using a screen reader. In such cases, an AI-generated summary of the text can help. 

Summaries break down lengthy text into smaller chunks that are easier to understand. Salesforce has been experimenting with and developing technology for this purpose, leading to advancements in natural language processing (NLP). Their software uses reinforcement learning (RL) and contextual word-generating models to provide accurate text summaries that are easy to read.

Lip reading

Algorithms for automating lip-reading are very similar to captioning algorithms and can be helpful to people with hearing impairment.

DeepMind by Google is at the forefront of this AI technology. After analyzing over 5,000 hours of TV shows, DeepMind can identify specific lip movements for particular words with an accuracy rate of 46.8%, allowing it to convert speech into text in real-time.

Project Euphonia by Google is another technology in this realm. It uses artificial intelligence to help decode human speech. Persons who have atypical speech can provide voice samples to help the AI learn to understand more diverse ways of speaking. This data can be used to advance speech recognition, giving more people the option to use voice-controlled technology moving forward.

User navigation

Keyboard traps — when a user using the keyboard is unable to shift focus away from an interactive control or element using only the keyboard — can significantly hamper the user experience. AI tools can be used to simulate user behavior to test for, identify, and prevent keyboard traps and other navigation issues.

Regression testing

Whenever any software is updated it could impact compliance with accessibility standards. Regression testing will ensure everything works as planned even when new changes are introduced. Machine learning helps make regression testing easier by using past data to automatically identify and address changes without manual intervention.

Artificial intelligence is helping transform the way people access information and interact with others in a digital world. Developers should continue to explore ways to use AI and machine learning-based solutions to make their digital content more accessible and inclusive. 

 

PDF Remediation Software - Free Online Event!

Join us on Tuesday, April 9th, at 1 PM ET for a free online event to explore how to evaluate and select PDF remediation software for your business. Click here to learn more about this event and to register.

Click here to see our Events Calendar.

Accessibility.com's 2024 events will utilize the Zoom Events platform, offering a virtual expo hall for attendees to meet with prospective vendors. If your company is interested in being part of the expo hall, don't hesitate to get in touch with Amanda@Accessibility.com.

Vendor Directory

Accessibility.com offers the premier impartial listing of digital accessibility vendors.  Search for products and services by category, subcategory, or company name.  Check out our new Vendor Directory here.

Comments