Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

AI integrations #2

Open
mgifford opened this issue May 11, 2021 · 3 comments
Open

AI integrations #2

mgifford opened this issue May 11, 2021 · 3 comments

Comments

@mgifford
Copy link

You've integrated what you plan to accomplish here - https://radix.ai/cases/putting-the-ai-into-accessibility/#the-briefing

Leveraging axe or HTML Code Sniffer will get you some of the way. I'm trying to learn how you plan to incorporate machine learning into this mix.

Tools like this can easily scan domains https://github.com/MSU-NatSci/DomainAccessibilityAudit

Extending this though to look for those issues which axe can't easily catch is a challenge.

Would be interesting to see how your efforts compare with https://www.evinced.com/

@bdeclerc
Copy link

We've implemented a domain scanner based on HTML Codesniffer here : https://github.com/openfed/AccessibilityCheckServer
One concern right now on transferring the AI integrations into a production environment is that they are quite resource-heavy to perform, orders of magnitude more CPU/Memory required than for the regular javascript-based checks in Codesniffer.

So further optimisation will almost certainly be required after this proof-of-concept project to make it available for production use (either that or reserving an entire Amazon datacenter perhaps...)

@bdeclerc
Copy link

Thanks for the evinced link BTW, looking into it.

@mgifford
Copy link
Author

Thanks @bdeclerc I'll check out your scanner. AI is a real challenge as most approaches require a lot of data as well as considerable processing power to be effective. Hopefully it won't take a whole datacentre, but yes :)

I think that currently the goal of AI is to help find automated means to find accessibility errors that are caught by people. It is probably a better goal to help find things that people aren't particularly good at. I'd love AI to start looking at dividing up and organizing errors so that they can be better addressed.

Can we find errors that are caused by the theme, a module or the CMS? Is it possible to identify common problems across a department or agency, that might be due to some out-of-date training material? Can we limit the errors we show to teams to those they have the power to change? Is it possible to link up to existing issue queues in modules/themes used to search for related issues? Are we able to see what changes in human behavior are making an impact?

In anycase, we are looking forward to seeing if there are elements that can be replicated on this side of the pond.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants