Deque Brings ML to Accessibility
March 10, 2020
an industry first, Deque has successfully integrated Machine Learning
technology to perform powerful visual analyses within axe Pro’s
automated and intelligent guided testing, which significantly reduces
the amount of manual work required to identify and fix accessibility
issues. Catching these issues quickly and easily is a crucial step to
ensure that websites and apps are accessible to all people, including
those with disabilities.
“Much of accessibility testing involves determining whether digital
content is accurately conveyed to assistive technologies and the users
who rely on them to access that content,” comments Preety Kumar, CEO,
Deque Systems. “By leveraging Machine Learning technology, we’ve
continued to automate many legacy manual testing efforts, drastically
reducing testing costs and making better use of a developer’s time.”
Learning - How It Works: axe Pro’s guided testing tool leverages
Computer Vision, which pulls semantic data from digital images, to
analyze the visual content of an HTML page’s interface elements and
their relationships to one another. The visual text is then extracted
from the page and fed into axe Pro’s guided testing tools, which then
performs an accessibility analysis in a fully automated fashion. This
greatly reduces the manual work typically required of such an analysis.
"Amazon, Google, Facebook and others have made huge advances in Machine
Learning and Computer Vision technologies, which we are leveraging to
apply to accessibility testing," says Dylan Barrell, CTO, Deque Systems.
"When added to our proven technology, and our domain-specific data, we
believe this first step constitutes a breakthrough that significantly
reduces the amount of manual testing. We will continue to use these
techniques to increasingly and steadily offer more automation and time
savings far into the future.”
Deque continues to work on improvements and additions to the axe Pro
beta. More exciting guided tests, new features and new Machine
Learning-powered tools will be added in coming weeks.