W3C home > Mailing lists > Public > w3c-wai-gl@w3.org > January to March 2018

Re: Finding agreement on common purpose

From: David MacDonald <david100@sympatico.ca>
Date: Tue, 16 Jan 2018 10:47:34 -0500
Message-ID: <CAAdDpDbypKnzAGAPds-tgAmq6QkwG9WCGtZ5D18nC9p4H_mzbQ@mail.gmail.com>
To: "Patrick H. Lauke" <redux@splintered.co.uk>
Cc: WCAG <w3c-wai-gl@w3.org>
> How? Machine learning?

We've been hearing about mapping documents for months and never seen one...
that's what I'm talking about...

Anyway... I'm ready to drop AccName being sufficient and tell everyone to
use autofill... but I know there will be a lot of authors and their
companies bummed out to have to do that on a bunch of forms that have a
programmatically determinable obvious name.

David MacDonald

*Can**Adapt* *Solutions Inc.*

Tel:  613.235.4902



GitHub <https://github.com/DavidMacDonald>

www.Can-Adapt.com <http://www.can-adapt.com/>

*  Adapting the web to all users*
*            Including those with disabilities*

If you are not the intended recipient, please review our privacy policy

On Tue, Jan 16, 2018 at 10:41 AM, Patrick H. Lauke <redux@splintered.co.uk>

> On 16/01/2018 15:31, David MacDonald wrote:
>> It was dropped a while ago due to internationalisation issues,
>> I agree languages other than English would be precarious. However, if an
>> English author wants to use the appropriate "purpose" as the label, I don't
>> think they should be punished by also having to add redundant metadata.
> It's not "redundant". It needs to be part of a restricted set of
> machine-readable tokens to be...well...machine-readable. Otherwise, you
> enter the realm of heuristics again (where hypothetical tools that
> hypothetically do something useful based on a control/element's purpose
> need to guess what the purpose is based on natural language processing or
> similar).
>  >it would limit the terms that people could use in labels, which was not
>> acceptable.
>> No, it would cause them to have to add metadata to provide that proper
>> "purpose" term.
> Alastair is right: authors would need to carefully craft their
> labels/accnames to make sure they trigger the right heuristics (and then
> that would require testing in all sorts of user agents, to make sure their
> natural language processing reacts consistently).
> I think the AT should be able to figure this out...
> How? Machine learning?
> P
> --
> Patrick H. Lauke
> www.splintered.co.uk | https://github.com/patrickhlauke
> http://flickr.com/photos/redux/ | http://redux.deviantart.com
> twitter: @patrick_h_lauke | skype: patrick_h_lauke
Received on Tuesday, 16 January 2018 15:48:00 UTC

This archive was generated by hypermail 2.4.0 : Thursday, 24 March 2022 21:08:21 UTC