Re: Joint mtg followup re XAUR and Avatars

Hi Janina and all,

I've an update on this, using your suggestions. I've added an edit that 
I think covers the informative need, and what we are trying to indicate 
but doesn't go into the numbers of facial muscles like you suggested. I 
think this make it pretty unambiguous though, that these are serious 
potential requirement if quality signing avatars are to fly. There are 
now two notes, and they read as follows:

Note

Currently, it is not possible to provide an accurate live interpretation 
via a signing avatar. In general, animated or digital signing avatars 
should be avoided as users find them less expressive than recorded video 
of humans who can convey the natural quality and skill provided by 
appropriately trained and qualified interpreters and translators. 
Therefore, uses of signed avatars should rely only on pre-recording of 
'real people' who are trained and qualified interpreters and 
translators. See the concerns expressed by the WFD and WASLI 'Statement 
on Use of Signing Avatars'. [wfd-wasli 
<http://raw.githack.com/w3c/apa/xaur-signing-avatar/xaur/index.html#bib-wfd-wasli>] 


Note

However, we note this is an emerging field and exploration is encouraged 
to ensure the future development of quality signing avatars. For 
example, this could be via building a signing avatar that both provides 
a face with fully functioning muscular variables and can successfully 
parse the nuances of vocal expression and meaning.


http://raw.githack.com/w3c/apa/xaur-signing-avatar/xaur/index.html#signing-videos-and-text-description-transformation

Please do let me know what you think of these edits, ta.

Josh

> Joshue O'Connor <mailto:joconnor@w3.org>
> Wednesday 26 May 2021 13:44
> Looks like a good addition Janina. I've edited it, as follows and 
> pushed this change to the branch:
>
>> Currently, it is not possible to provide an accurate live 
>> interpretation via a signing avatar. Animated or digital signing 
>> avatars should be avoided as users find them less expressive than 
>> recorded video of humans who can convey the natural quality and skill 
>> provided by appropriately trained and qualified interpreters and 
>> translators.  Therefore uses of signed avatars should rely only on 
>> pre-recording of 'real people' who are trained and qualified 
>> interpreters and translators. See the concerns expressed by the WFD 
>> and WASLI 'Statement on Use of Signing Avatars'. [[wfd-wasli]] 
> Looks good?
>
> Thanks
>
> Josh
>
>
>
>> Janina Sajka <mailto:janina@rednote.net>
>> Wednesday 26 May 2021 13:02
>> Trying a few edits ...
>>
>> Signing avatars should be avoided as users find them less expressive
>> than recorded video.
>>
>> Developers who wish to work on perfecting the use of avatars should be
>> certain to model human facial expression using all 44 muscles in the
>> human face when creating signing avatars, as facial expression is a key
>> component of effective sign language communication.
>>
>> Joshue O'Connor <mailto:joconnor@w3.org>
>> Wednesday 26 May 2021 12:34
>> Hi Janina,
>>
>> Based on the discussion last week I've updated the Signing avatar 
>> user need and requirement with some suggested text:
>>
>> <suggested update>
>>
>> 4.10 Text description transformation
>> User Need 10: A deaf or hard of hearing person, for whom English or 
>> any other written language, may not be their first language and may 
>> have a preference for signing of text alternatives or equivalents.
>> REQ 10a: Allow text, objects or item descriptions to be presented to 
>> the user via a signing avatar.
>> NOTE
>> Any signing avatar should convey the natural quality and skill 
>> provided by appropriately trained and qualified interpreters and 
>> translators. There have been concerns expressed by the WFD and WASLI 
>> 'Statement on Use of Signing Avatars' as machine translations have 
>> yet to emulate the human ability in creating a live interpretation 
>> (spoken or signed). Currently, it is not possible to provide an 
>> accurate live interpretation via a signing avatar. Therefore uses of 
>> signed avatars should rely on pre-recording of 'real people' who are 
>> trained and qualified interpreters and translators. [wfd-wasli]
>> </suggested update>
>>
>> https://raw.githack.com/w3c/apa/xaur-signing-avatar/xaur/index.html
>>
>> Thoughts and comments welcome, thanks
>>
>> Josh
>>
>> Janina Sajka <mailto:janina@rednote.net>
>> Wednesday 19 May 2021 16:17
>> I thought I should say for the record what I raised my hand to say when
>> we ran out of time ...
>>
>> Regarding avatars ...
>>
>> I was highly impressed by the detailes that emerged about why avatars
>> tend to fail SL users. I'm thinking we should capture a high level
>> description of what would be required to create a successful avatar in
>> the XAUR by way of answering any engineering interest in moving to their
>> use prematurely.
>>
>> I believe the explanation is that SL captures far more than the words
>> which are captured in a text transcript of what's being said. SL
>> attempts to communicate more of the conversation than just the verbal
>> language content we've learned to capture with paper and ink.
>>
>> Facial expression -- there are some 43 muscles that control facial
>> expression, though if one googles this question the answers vary, 43,
>> 42, 33 ...
>>
>> Implication: anyone building a signing avatar should provide a face and
>> 43 functioning muscular variables.
>>
>> Similarly, there's the challenge to understand the nuance of vocal
>> expression. Consider the word "O:"
>>
>> O (as in startled surprise)
>> O? (as in really?)
>> O (as in oops, which sometimes comes out as "o, o)
>>
>> There are more for just this one word, but I believe I've made my point.
>>
>> If we take this tack we avoid a perscription against engineering
>> development and supplant it with the far more meaningful challenge of
>> what it takes to design a satisfying avatar.
>>
>> Thoughts?
>>
>> Janina
>>
>>
>
> Janina Sajka <mailto:janina@rednote.net>
> Wednesday 26 May 2021 13:02
> Trying a few edits ...
>
> Signing avatars should be avoided as users find them less expressive
> than recorded video.
>
> Developers who wish to work on perfecting the use of avatars should be
> certain to model human facial expression using all 44 muscles in the
> human face when creating signing avatars, as facial expression is a key
> component of effective sign language communication.
>
> Joshue O'Connor <mailto:joconnor@w3.org>
> Wednesday 26 May 2021 12:34
> Hi Janina,
>
> Based on the discussion last week I've updated the Signing avatar user 
> need and requirement with some suggested text:
>
> <suggested update>
>
> 4.10 Text description transformation
> User Need 10: A deaf or hard of hearing person, for whom English or 
> any other written language, may not be their first language and may 
> have a preference for signing of text alternatives or equivalents.
> REQ 10a: Allow text, objects or item descriptions to be presented to 
> the user via a signing avatar.
> NOTE
> Any signing avatar should convey the natural quality and skill 
> provided by appropriately trained and qualified interpreters and 
> translators. There have been concerns expressed by the WFD and WASLI 
> 'Statement on Use of Signing Avatars' as machine translations have yet 
> to emulate the human ability in creating a live interpretation (spoken 
> or signed). Currently, it is not possible to provide an accurate live 
> interpretation via a signing avatar. Therefore uses of signed avatars 
> should rely on pre-recording of 'real people' who are trained and 
> qualified interpreters and translators. [wfd-wasli]
> </suggested update>
>
> https://raw.githack.com/w3c/apa/xaur-signing-avatar/xaur/index.html
>
> Thoughts and comments welcome, thanks
>
> Josh
>
> Janina Sajka <mailto:janina@rednote.net>
> Wednesday 19 May 2021 16:17
> I thought I should say for the record what I raised my hand to say when
> we ran out of time ...
>
> Regarding avatars ...
>
> I was highly impressed by the detailes that emerged about why avatars
> tend to fail SL users. I'm thinking we should capture a high level
> description of what would be required to create a successful avatar in
> the XAUR by way of answering any engineering interest in moving to their
> use prematurely.
>
> I believe the explanation is that SL captures far more than the words
> which are captured in a text transcript of what's being said. SL
> attempts to communicate more of the conversation than just the verbal
> language content we've learned to capture with paper and ink.
>
> Facial expression -- there are some 43 muscles that control facial
> expression, though if one googles this question the answers vary, 43,
> 42, 33 ...
>
> Implication: anyone building a signing avatar should provide a face and
> 43 functioning muscular variables.
>
> Similarly, there's the challenge to understand the nuance of vocal
> expression. Consider the word "O:"
>
> O (as in startled surprise)
> O? (as in really?)
> O (as in oops, which sometimes comes out as "o, o)
>
> There are more for just this one word, but I believe I've made my point.
>
> If we take this tack we avoid a perscription against engineering
> development and supplant it with the far more meaningful challenge of
> what it takes to design a satisfying avatar.
>
> Thoughts?
>
> Janina
>
>


-- 
Emerging Web Technology Specialist/Accessibility (WAI/W3C)