Silver Task Force & Community Group

04 Mar 2022


Lauriat, Daniel, Jennifer, Makoto, maryjom, ToddL, KimD, Wilco, janina, Rachael, SuzanneTaylor


<Lauriat> https://www.w3.org/WAI/GL/task-forces/silver/wiki/Scribe_List

<scribe> scribe: maryjom

Warning: clocks changing soon

Natural Language of Text

<dmontalvo> https://docs.google.com/document/d/1-c96lQsQq1jgdDKo4C9GaWEj7cVoxDdCvtLPvqCOBBE/editeading=h.y5p1v6vmgi1x

dm: Joint Silver and accessibility conformance testing groups have been working on this.

<Lauriat> WCAG 3: Pronunciation of Text https://docs.google.com/document/d/1-c96lQsQq1jgdDKo4C9GaWEj7cVoxDdCvtLPvqCOBBE/edit#heading=h.y5p1v6vmgi1x

dm: We are continuing to do research on when users use language change and when they don't.
... Users typically want the language to change by them changing settings themselves.
... We changed to use the term "speech synthesizers"
... The user is responsible for changing the braille table themselves using braille profiles.
... Latin pronunciations are different by locality, so it is difficult to know the right pronunciation of Latin.
... So we are working on how to handle these kinds of unique situations.
... There's a second document that came from our internationalization work.
... We are currently writing definitions of "blocks of text".
... We are working on what are or are not considered "blocks of text".
... We are seeking your feedback and possibly agreement.
... Looking for feedback on wording of the outcome. Want to know if we're going in the right direction.
... Want feedback on the level of granularity of "block of text" and how we are defining it.
... Don't think we should be requiring changes of language at a word level.

<janina> ~mute broken, come back to me

js: I think we need to do more coordination across W3C groups. Not sure block level is the lowest level that is acceptable.
... In APA, many are coming from an educational background needing language support even where there may not be lang packs.
... In Greek, if your a scholar of that era (Classical greek) and we don't have enough language distinction.
... We don't want to have competing guidance since there are different groups interested in this space have different thoughts.

dm: This work is coming from the ACT rules work. It would be good to learn your use cases and examples to inform the work.

wf: We are struggling to find research related to these topics. Difficult to figure out when changes of language are absolutely necessary.
... If you have research, please point us to that. We don't have data that guides us in what to require.

sl: We could have edge cases defined, and also have guidance that generally is correct in 80% of cases.

js: What we are developing in the pronunciation spec is to provide authoring level guidance.
... It may help solve it.

dm: At a more granular level than simply blocks there are a lot of nuances to make it difficult to make the requirement objective enough.

js: I think maybe we need to involve the internationalization group, because standard language identifiers and they only address spoken language.
... They don't cover TTS.
... We may need their help to define. Chaucer is pronounced differently than EE Cummings.

mu: In Japanese there are Kanji characters that are pronounced differently. We have luby that helps define how the Kanji is pronounced.
... Will this outcome cover the Japanese pronunciation by AT?

dm: No, we are not covering AT pronunciation rules. But if there's author-controlled markup it would cover.

wf: We'll be in touch to make sure Japanese concerns are considered.

<Lauriat> chrome://flags/#enable-experimental-accessibility-language-detection

sl: Chrome has some experiments to enable. See above link.

<Lauriat> chrome://flags/#enable-experimental-accessibility-language-detection-dynamic

sl: There's some language detection. I have not tried them. Two flags available to try them out. Not detecting language below block level.
... Markup of language is exposed up in the accessibility tree and they don't do that. It detects at the block level.

dm: Thanks for that, will try it.
... Edge cases: Should a French word in the middle of an English sentence be marked up vs. a French sentence in the middle of an English paragraph.

js: One of the ARIA failures is that it doesn't support Braille well.
... Frequently a word in the middle of the sentence can break down. Capital of South Dakota is Pierre and it doesn't use the French pronunciation.
... So this kind of situation that will cause mispronunciations.

sl: What will help is documenting use cases that support the direction you take and will help everyone understand the goal for splitting.
... Examples are really helpful.
... Wondering if there are cases are important for voice commands, trying to interact with the page or select some text using voice.

<Zakim> Lauriat, you wanted to respond to Wilco's question about splitting pronunciation and Braille and to respond to Wilco's question about splitting pronunciation and Braille, and ask

js: It may get taken up as part of the Natural Language work covered in their use cases. I belive it's addressed there.

<Zakim> MichaelC, you wanted to suggest an ¨average person¨ standard

js: Feel free to comment on that, I'll try to find pointers.

mc: We need to think about what the average person would do when mispronouncing words. We all mispronounce words and they are still understood.
... We shouldn't overprescribe.

<Wilco> +1

<janina> Pronunciation Gap Analysis & Use Cases:

<janina> https://www.w3.org/TR/pronunciation-gap-analysis-and-use-cases/

<MichaelC> I should note that I made my comment as a non screen reader user, and my view of ¨acceptable mispronunciation¨ may be different from others´, please challenge if needed

dm: I found one case in Braille where language changes, but mostly it is up to the user to change the labels.

<janina> Natural Language AUR: https://www.w3.org/WAI/APA/task-forces/research-questions/wiki/Main_Page#Natural_Language_Interface_Accessibility_User_Requirements

dm: We might want to add something about Braille and how the user preferences work. Let us know if we're missing something.

sl: Practical examples will also help for people to understand and give feedback on that.
... I have been traveling and the pronunciation changed automatically in GPS when in Canada vs. the US. Went into a strong Canadian accent.
... There seem to be a lot of overlap between pronunciation and dictionaries.

js: The ARIA implementers are proposing a dictionary.

Warning: clocks changing soon

Overview of test & scope structure

<Lauriat> Exploratory test breakdown https://docs.google.com/presentation/d/1lU94dPLGa7ZccOKuFvtM44fVTT0tqLdrxFlvy-XwL04/edit#slide=id.g117d91f9d2d_0_0

<Jennifer> * I wish he didn't mute, that would be interesting. ;)

sl: This is a follow-up from a previous discussion. This is a working document.
... From process and scope sub-group working through how to articulate kinds of tests and their scope.
... Took WCAG 1.1.1 just for alt text. Broke it into the smallest unit view, process, etc.
... This has no context yet. This is not something we're using for migration of guidelines from WCAG 2.0 to 3.0.
... The smallest unit is a concrete label - very yes or no. Is there an alt attribute?
... Measure of length of alt text. Conditional tests that you write steps to follow to make a subjective judgement call.
... So one checks alt text is meaningful, or concise.
... Then there's context of pages under test, e.g. managing a new site with articles that have hero images.
... You can write tests to ensure the alt text is suitable for the context where the image is used.
... You could write automated tests that check that it is the correct usage of an image.
... Defining use cases and target lengths would be helpful in this work.

<Rachael> Example would be the organization conducts a plain language review

sl: An organization can write a protocol for alternative text - to show you are more likely to have correct alt text.

wf: What problem are we trying to solve with this model?

sl: To add some structure to conversations of how to test things. There's a range, and it makes it easier to think through the scenarios.

<janina> Do I understand greater probability for what is a qualitative test?

sl: e.g. We could have a range of scores from fundamental tests like is the alt meaningful. How do you build up to a judgement alt is appropriate?
... This table represents several concepts: process and scope of test, aggregating, overall level of conformance, and protocols.
... So you could not only judge alt text itself, but also the processes used.

r: We all tend to have difficulty with the approach of increasing subjectivity while still addressing conformance.
... What can we get to be pass/fail. By breaking this into this table we can think through that key question.

sl: We could potentially use this in the template for thinking through tests as we work on the requirements in WCAG 3
... What are your thoughts around adding consistent navigation. Writing tests for meeting the bar for consistent navigation, how would you structure tests?

suzanne: This is helpful. The text may be written by someone who isn't a subject matter expert.

<janina> +1 to getting authoring of alt text into the process for subject matter experts

suzanne: This chart, if someone's writing a blog they may not take as much care in writing the alt text. But other cases it may be scrutinized more carefully.

<janina> CSS Media Queries now gives us control over how much motion -- ostensibly

sl: Another example is animations. If it is interactive, how severe/pronounced is that motion. Protocol could be to make sure you have something in your design system on animation.
... Is there something that is missing here?

suzanne: I've seen cases where 2 buttons individually pass WCAG, but because they were next to each other there was too much motion.
... Alone, one button wasn't a problem, together, it was.

suzanne. This helps looking also at a more meta/process level so you can catch this sort of thing.

wf: This seems to be making things more complex, but willing to have this explored further.

sl: Maybe we should separate out the protocol, to simplify the table. This could help make the table simpler with 3-4 concepts to think about.

Summary of Action Items

Summary of Resolutions

[End of minutes]

Minutes manually created (not a transcript), formatted by David Booth's scribe.perl version 1.200 (CVS log)
$Date: 2022/03/04 16:03:09 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision VERSION of 2020-12-31
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00)

Succeeded: s/works/words/
Default Present: Lauriat, Daniel, Jennifer, Makoto, maryjom, ToddL, KimD, Wilco, janina, Rachael, SuzanneTaylor
Present: Lauriat, Daniel, Jennifer, Makoto, maryjom, ToddL, KimD, Wilco, janina, Rachael, SuzanneTaylor
Found Scribe: maryjom
Inferring ScribeNick: maryjom

WARNING: No meeting chair found!
You should specify the meeting chair like this:
<dbooth> Chair: dbooth

WARNING: No date found!  Assuming today.  (Hint: Specify
the W3C IRC log URL, and the date will be determined from that.)
Or specify the date like this:
<dbooth> Date: 12 Sep 2002

People with action items: 

WARNING: Input appears to use implicit continuation lines.
You may need the "-implicitContinuations" option.

WARNING: IRC log location not specified!  (You can ignore this 
warning if you do not want the generated minutes to contain 
a link to the original IRC log.)

[End of scribe.perl diagnostic output]