W3C

- DRAFT -

Accessible Platform Architectures Working Group Teleconference

12 Aug 2020

Attendees

Present
jasonjgw, LisaSeemanKest_, janina, SteveNoble, nicolocarp
Regrets
Scott, Judy
Chair
jasonjgw
Scribe
SteveNoble

Contents


<LisaSeemanKest_> can someone msg me the zoom link?

<LisaSeemanKest_> can someone msg me the zoom link?

Miscellaneous updates.

<janina> Chair: Jason

Jason: machine learning can aid in setup of personalization based on needs and preferences

Lisa: machine learning applications which have connections to COGA

Josh: good idea to raise a flag for this space at the workshop

Jason: research by machine learning specialists at Google about technologies that can support form-filling, etc.
... text-summarizing is another angle

Lisa: advice like making sure a button looks like a button...machine learning can only go so far to make this identification
... AI-training is dependent upon the body of examples it has trained on
... Machine learning does not do away with the need for user testing
... The context of the purpose of the button is much more significant then the fact it is a button
... a major problem comes when trying to map concepts to language
... Instead of words, symbols which map to concepts provide a rich vocabulary
... One connection might be to sign-language avatars
... machine learning could populate the correct symbols
... example of problems with automated language translations which may be more familiar
... being able to link up to a familiar interface the individual is comfortable with
... Symbols can help where language translation in words may fail

<Zakim> Joshue_, you wanted to say the wiki page could be updated https://www.w3.org/WAI/APA/wiki/MLWorkshop2020

Josh: Would like to flag some ideas in the workshop proposal
... If we have something within the next week we should be able to proceed to submit

Janina: The problem of ACT-like binary rules which do not fit so well in this realm

Lisa: It is tricky to say whether or not image identification of buttons and such things are testable

Janina: Machine learning may enhance experience for a wide sector of the population

Jason: will put this back on the agenda when we have the proposal in hand

Lisa: will try to do something in the next week

Jason: if we can introduce some additional possibilities into other things that they are planning

<Joshue_108> Lisa here is the speakers info page for the ML workshop https://www.w3.org/2020/06/machine-learning-workshop/speakers.html

Lisa: one problem is that developers are not considering the COGA population
... Individuals with various disabilities often fail to be included in the corpus that AI is being trained

Jason: mapping workshop?

Lisa: COGA may be looking at the mapping workshop

Josh: Will let the workshop planners know what we are planning

<Zakim> janina, you wanted to talk about ACT rules as we might explain them in the workshop

Media synchronization requirements.

Janina: lack of a specification of what synchronization means
... use case lip reading - when video and audio are out of sync even slightly can be a major problem, though captions not as much
... would need to define a level of tolerance +/- which will vary between use case

Jason: will need to do some research to see what is in the literature
... synced highlighting of text and speech - word level

<Joshue_108> SN: There is the example of highlighting with Math text to speech.

<Joshue_108> Math player is good at this also.

<jasonjgw> Steve notes highlighting synchronized with spoken mathematics, as in the MathPlayer implementation.

<jasonjgw> The structural components of the mathematical expression are highlighted as they are spoken

<Zakim> Joshue_, you wanted to talk about scrubbing in XR and syncronisation

Jason: consideration of what is considered adeqyate for each user case
... will bring this back next week

Josh: about half-way thruogh on updating XAUR
... will work with Lisa on the workshop abstract

Jason: next week - synchronization, plus other updates

Summary of Action Items

Summary of Resolutions

[End of minutes]

Minutes manually created (not a transcript), formatted by David Booth's scribe.perl version (CVS log)
$Date: 2020/08/12 14:00:46 $

Scribe.perl diagnostic output

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

Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00)

Default Present: jasonjgw, LisaSeemanKest_, janina, SteveNoble, nicolocarp
Present: jasonjgw LisaSeemanKest_ janina SteveNoble nicolocarp
Regrets: Scott Judy
No ScribeNick specified.  Guessing ScribeNick: SteveNoble
Inferring Scribes: SteveNoble
Found Date: 12 Aug 2020
People with action items: 

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]