Re: An approach suggestion - ways to test qualitative content

Hey Alistair,

Thanks for putting this together. I think you and I are roughly on the same
line on how such testing could work. I'd be interested to learn your
thoughts on testing in context. Take your two dogs image example. "Two dogs
running after a ball" may be a pretty good description for it, but "Bob and
Jack playing during vacation" may very well be a good alternative too,
depending on the context. How do you enable testing such things, without
making the test extremely permissive?

I think having a big dataset to train machine learning algorithms on for
testing accessibility would be a great boon to the world. I'm not sure how
we'd go about developing such a thing, or if that is necessarily connected
to the development of WCAG 3 itself.


On Wed, Oct 20, 2021 at 12:07 PM Alistair Garrison <
alistair.garrison@accesseo.co.uk> wrote:

> Dear All,
>
> I’d like to present the following concept paper, entitled "Measuring Alt
> Text Quality”.
>
>
> https://docs.google.com/document/d/1IVCPcfyfnjm2RJRvgV5pbcQLwzxXKHw66V7R-OW3wiE/edit?usp=sharing
>
> Apologies that it is in Google Docs.
>
> It is, of course, rough and ready as I just wanted to get something
> written down to start the ball rolling.  It is not perfect (and might be
> subject to minor changes up until 30th Oct 2021), but it has stimulated a
> lot of thought in my own mind, and I hope it does in yours.  It also shows
> that metrics (such as SPICE), which are available in the broader AI/ML/CV
> research world, might prove useful when considering things we are looking
> to solve in the field of web accessibility.
>
> Very best regards
>
> Alistair
>
> Alistair Garrison
> CEO / Founder Accesseo Limited
>
>
>
>
>
>
>
>
>
>
>
>

-- 
*Wilco Fiers*
Axe-core & Axe-linter product owner - WCAG 3 Project Manager - Facilitator
ACT Task Force