E-mail List Archives

Search E-mail List Archives for

Number of posts in this thread: 9 (In chronological order)

From: Tomlins Diane
Date: Tue, Jun 27 2017 2:06PM
Subject: SC 1.2.2 and YouTube's auto captions
No previous message | Next message →

I have some folks who want to argue that YouTube's automatic captions will pass the 1.2.2 Criteria - "look, it has captions!". While I know relying on YT's auto captions is just a bad idea since they are often so wrong, DOES it actually pass the criteria if the CC is available? Or does it fail regardless, as per this example?

https://www.w3.org/TR/2016/NOTE-WCAG20-TECHS-20161007/F8

I'd like to make sure I am armed with enough material to get them to understand that auto captions doesn't cut it.

Thanks!

Thanks!
Diane R Tomlins
HCA IT&S | Digital Media
Accessibility SME

From: John Foliot
Date: Tue, Jun 27 2017 2:36PM
Subject: Re: SC 1.2.2 and YouTube's auto captions
← Previous message | Next message →

Hi Diane,

To answer that question requires another question: what is your ultimate
goal here? To be in compliance with WCAG 2.0 for legal reasons, or to use
SC 1.2.2 Criteria to make video content truly accessible to those end-user
who require captions?

If your response is to meet the legal requirement, the entire Success
Criteria states:

*Captions are provided for all prerecorded audio content in synchronized
media, except when the media is a media alternative for text and is clearly
labeled as such.*
(source:
https://www.w3.org/TR/WCAG20/#media-equiv)


Sadly, there is nothing in that language that specifies *accuracy*, only
that captions must exist.

From a legal perspective, that would be at least one valid interpretation,
and since we don't actually have any legal precedent to fall back on here,
it has to remain as simply that for now: an interpretation.

True, the Technique you reference suggests that inaccurate captions fail to
meet the end-user need, but given the way that WCAG is written, all
Techniques (both Passing Techniques and Failure Techniques) are
non-normative
<https://english.stackexchange.com/questions/28882/meaning-of-non-normative>:
they are not actually part of the standard/specification (which has now
taken on an additional "legal" perspective).

When presented to a judge, we simply don't know how that judge will react.


The real focus then needs to be on the functional requirement (and not the
legal one), which is where the Technique you reference comes in, and that
takes a different strategy. I have found that focusing on the negative
impact on a Brand perception ("Look at this, how dumb can they be", or an
automated caption that mangles a product name, or a senior executive's
family name) all because the site didn't go the additional way and ensure
accurate captions has way more impact and resonance.

In other words, sell the upsides, and not the downsides.

But to the question of "are YouTube's automatic captions enough?" the
actual answer is, legally, we just don't know for sure.

JF
--
John Foliot
Principal Accessibility Strategist
Deque Systems Inc.
= EMAIL ADDRESS REMOVED =

Advancing the mission of digital accessibility and inclusion

From: Jared Smith
Date: Tue, Jun 27 2017 2:59PM
Subject: Re: SC 1.2.2 and YouTube's auto captions
← Previous message | Next message →

John Foliot wrote:

> Sadly, there is nothing in that language that specifies *accuracy*, only
> that captions must exist.

While there's not a defined accuracy level (e.g., 95%), the WCAG
normative definition for "captions" -
https://www.w3.org/TR/WCAG20/#captionsdef - does prescribe that the
captions provide "information needed to understand the media content"
and that "captions convey not only the content of spoken dialogue, but
also equivalents for non-dialogue audio information".

By my interpretation, if the captions don't allow the user to
understand the media content or don't convey the content of the spoken
dialogue and non-dialogue audio, then they would not be conformant.
Measuring this definitively is a bit difficult, though it's clear that
many (most?) YouTube auto-captions would not align with this
definition.

Jared

From: Tomlins Diane
Date: Tue, Jun 27 2017 3:04PM
Subject: Re: SC 1.2.2 and YouTube's auto captions
← Previous message | Next message →

Thanks John, great points..

>To answer that question requires another question: what is your ultimate goal here? To be in compliance with WCAG 2.0 for legal reasons, or to use SC 1.2.2 Criteria to make video content truly accessible to those end-user who require captions?

And the answer is not cut and dried either, unfortunately. The *ultimate* goal is to make videos truly accessible for those who require captions and I won't stop pushing for that.

The near-term answer - is that we need to meet SC 1.2.2 for compliance to the standard. Since accuracy sadly doesn't fail the criteria (yet), then technically we'd be covered from that perspective.

I want to make it right the first time around, but I probably won't win that argument today - and just insist that proper captions are added in an upcoming production push. :-/

Diane R Tomlins
HCA IT&S | Digital Media
Hi Diane,

From: Angela French
Date: Tue, Jun 27 2017 3:20PM
Subject: Re: SC 1.2.2 and YouTube's auto captions
← Previous message | Next message →

Maybe this would be a smart aleck answer for them but ask them if html pages with misspelled text all over them is okay? I suspect the answer would be no.

Angela French
SBCTC

From: Patrick H. Lauke
Date: Tue, Jun 27 2017 3:25PM
Subject: Re: SC 1.2.2 and YouTube's auto captions
← Previous message | Next message →

On 27/06/2017 21:59, Jared Smith wrote:
[...]
> Measuring this definitively is a bit difficult, though it's clear that
> many (most?) YouTube auto-captions would not align with this
> definition.

As a side note, I found recently that YouTube audio captions for
presentations (from http://www.inclusivedesign24.org/) were often more
accurate than the live captions we received (and are currently
uploading). So this definitely depends (and I have a feeling that the
accuracy of these auto captions has increased quite dramatically, no
doubt due to advances in speech recognition solutions).

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

From: John Foliot
Date: Tue, Jun 27 2017 3:28PM
Subject: Re: SC 1.2.2 and YouTube's auto captions
← Previous message | Next message →

> By my interpretation, if the captions don't allow the user to understand
the media content or don't convey the content of the spoken dialogue and
non-dialogue audio, then they would not be conformant.

I don't think Jared and I are far apart (we usually aren't), and I don't
disagree with his interpretation.

The larger issue however is that we don't have a legally defined
interpretation, which means we have to rely on logic and intuition, and our
understanding of the problem space. That's relatively easy for us as SME's
(beginner or experienced - if you are reading this I will call you a
Subject Matter Expert).

My only point is that there are many ways to look at this: legally,
practically, and holistically, and each perspective will likely land you at
a different response. I'm simply suggesting that one legal interpretation
of the actual Success Criteria would not include 'accuracy' as one of the
requirements, at least not explicitly, because that term and concept is not
mentioned in the normative Success Criteria. That's not great, but it is
accurate.

JF

On Tue, Jun 27, 2017 at 4:20 PM, Angela French < = EMAIL ADDRESS REMOVED = > wrote:

> Maybe this would be a smart aleck answer for them but ask them if html
> pages with misspelled text all over them is okay? I suspect the answer
> would be no.
>
> Angela French
> SBCTC
>
>

From: Tomlins Diane
Date: Wed, Jun 28 2017 7:51AM
Subject: Re: SC 1.2.2 and YouTube's auto captions
← Previous message | Next message →

Jared wrote:
> By my interpretation, if the captions don't allow the user to understand the media content or don't convey the content of the spoken dialogue and non-dialogue audio, then they would not be conformant.
Measuring this definitively is a bit difficult, though it's clear that many (most?) YouTube auto-captions would not align with this definition.

This was my take as well when reading the SC, since so many of the captions I see are so far off. Fortunately the one in question for this one application is not terrible, but it's not completely accurate either. Thankfully it's also fairly short and there's no medical terminology involved. Auto captioning is really bad with that and with anyone with an accent.

Thanks for all of your input! :)

Diane R Tomlins
HCA IT&S | Digital Media
Accessibility SME

From: Alastair Campbell
Date: Fri, Jul 07 2017 7:41AM
Subject: Re: SC 1.2.2 and YouTube's auto captions
← Previous message | No next message

Taking the wholistic view, I'd point out that auto-captions should be seen
as a helpful start on the process, not the finished product.

It used to be a complete pain in the proverbial to type out captions, put
them in XML, work out the timings etc.
I avoided it whenever possible, usually by avoiding doing video.

If you use youtube, you can login and update the caption text very easily,
this makes a huge difference overall. There are other services and methods,
but the room for excuses is getting smaller and smaller.

Cheers,

-Alastair