WebAIM - Web Accessibility In Mind

E-mail List Archives

Thread: Machine learning for accessibility tooling in Firefox

for

Number of posts in this thread: 4 (In chronological order)

From: Asa Dotzler
Date: Tue, Jul 23 2019 1:25PM
Subject: Machine learning for accessibility tooling in Firefox
No previous message | Next message →

Many web sites are inaccessible because they use custom page elements that
lack the accessibility of browser-native ones. You have probably run across
this in your surfing: a page that uses divs and JavaScript to create a
checkbox, for example. Mozilla wants to improve this by identifying
inaccessible custom elements in the Firefox developer tools and helping web
authors use standard ones—or helping them add the right ARIA roles to make
their custom elements accessible.

But identifying inaccessible custom elements is a challenge. The approach
we're investigating is to use machine learning, which requires many
examples. We would greatly appreciate your help in finding them, via this
survey

https://docs.google.com/forms/d/e/1FAIpQLSfkZ_Hq4UKUtJDynbaZsSdfZEDx3Un9Z7M-MBN4ofy7vOvhlQ/viewform

Thank you,
Asa Dotzler

From: Date: Tue, Jul 23 2019 1:50PM
Subject: Re: Machine learning for accessibility tooling in Firefox
← Previous message | Next message →

On 23/07/2019 20:25, Asa Dotzler wrote:
[...]

> But identifying inaccessible custom elements is a challenge. The approach
> we're investigating is to use machine learning, which requires many
> examples. We would greatly appreciate your help in finding them, via this
> survey
>
> https://docs.google.com/forms/d/e/1FAIpQLSfkZ_Hq4UKUtJDynbaZsSdfZEDx3Un9Z7M-MBN4ofy7vOvhlQ/viewform

Has Mozilla considered grepping the web for class name "cow paths"? I'd
wager that looking for things like "class="button", class="btn" and
variants would turn up a huge amount of data to work with.

If memory serves, this is one of the methods used to identify the cow
paths that were later paved as the header, footer, navigation elements
and such.


Léonie.



>
> Thank you,
> Asa Dotzler
> > > > >

--
@TetraLogical TetraLogical.com

From: Mallory
Date: Wed, Jul 24 2019 4:33AM
Subject: Re: Machine learning for accessibility tooling in Firefox
← Previous message | Next message →

Unfortunately many of the bad patterns I see are sites I've audited and that can fall under an NDA. However they are very often using open libraries. For example I recently did a site using <hi-button> everywhere. Are links to libraries also useful for this survey?

Sometimes the libraries' own docs will explain how do to something in an accessible manner but clearly the developers using the library just copy and paste or feel they don't have time to read 300+ pages and so try to treat the docs like a cookbook. Would these libraries still be useful in the survey?

cheers,
Mallory

On Tue, Jul 23, 2019, at 9:50 PM, Léonie Watson via WebAIM-Forum wrote:
>
> On 23/07/2019 20:25, Asa Dotzler wrote:
> [...]
>
> > But identifying inaccessible custom elements is a challenge. The approach
> > we're investigating is to use machine learning, which requires many
> > examples. We would greatly appreciate your help in finding them, via this
> > survey
> >
> > https://docs.google.com/forms/d/e/1FAIpQLSfkZ_Hq4UKUtJDynbaZsSdfZEDx3Un9Z7M-MBN4ofy7vOvhlQ/viewform
>
> Has Mozilla considered grepping the web for class name "cow paths"? I'd
> wager that looking for things like "class="button", class="btn" and
> variants would turn up a huge amount of data to work with.
>
> If memory serves, this is one of the methods used to identify the cow
> paths that were later paved as the header, footer, navigation elements
> and such.
>
>
> Léonie.
>
>
>
> >
> > Thank you,
> > Asa Dotzler
> > > > > > > > > >
>
> --
> @TetraLogical TetraLogical.com
> > > > >

From: Brandon Keith Biggs
Date: Wed, Jul 24 2019 5:06AM
Subject: Re: Machine learning for accessibility tooling in Firefox
← Previous message | No next message

Hello,
Yes, libraries are often the problem. Most of the React, Angular, and Vue
libraries I've seen are not accessible in many ways.
Thanks,

Brandon Keith Biggs <http://brandonkeithbiggs.com/>;


On Wed, Jul 24, 2019 at 3:33 AM Mallory < = EMAIL ADDRESS REMOVED = > wrote:

> Unfortunately many of the bad patterns I see are sites I've audited and
> that can fall under an NDA. However they are very often using open
> libraries. For example I recently did a site using <hi-button> everywhere.
> Are links to libraries also useful for this survey?
>
> Sometimes the libraries' own docs will explain how do to something in an
> accessible manner but clearly the developers using the library just copy
> and paste or feel they don't have time to read 300+ pages and so try to
> treat the docs like a cookbook. Would these libraries still be useful in
> the survey?
>
> cheers,
> Mallory
>
> On Tue, Jul 23, 2019, at 9:50 PM, Léonie Watson via WebAIM-Forum wrote:
> >
> > On 23/07/2019 20:25, Asa Dotzler wrote:
> > [...]
> >
> > > But identifying inaccessible custom elements is a challenge. The
> approach
> > > we're investigating is to use machine learning, which requires many
> > > examples. We would greatly appreciate your help in finding them, via
> this
> > > survey
> > >
> > >
> https://docs.google.com/forms/d/e/1FAIpQLSfkZ_Hq4UKUtJDynbaZsSdfZEDx3Un9Z7M-MBN4ofy7vOvhlQ/viewform
> >
> > Has Mozilla considered grepping the web for class name "cow paths"? I'd
> > wager that looking for things like "class="button", class="btn" and
> > variants would turn up a huge amount of data to work with.
> >
> > If memory serves, this is one of the methods used to identify the cow
> > paths that were later paved as the header, footer, navigation elements
> > and such.
> >
> >
> > Léonie.
> >
> >
> >
> > >
> > > Thank you,
> > > Asa Dotzler
> > > > > > > > > > > > > > >
> >
> > --
> > @TetraLogical TetraLogical.com
> > > > > > > > > >
> > > > >