Spoken Pronunciation Task Force Teleconference

19 December 2022


IrfanAli, NeilS

Meeting minutes

janina: it is important to figure out the next action items for this task force

janina: we can talk about our calendar.

Holiday calendar

janina: we are meeting on Monday which means we are not meeting on January 2nd but we will be meeting on 9th

janina: another holiday is Martin Luther on January 16. May be we will be meeting 23rd

janina: Proposing meeting on 9th and 23rd

plan for closing on technical direction

janina: we heard from WHATWG that they would entertain thing as native component. when we discussed the draft message, we talked if thats the direction if we want to go considering SR questions.

janina: we needed to have this discussion here within this group to ensure the direction

janina: we had discussion outside of the group at APA planning. It is heavy lift for SR to parsing. we find that this is optimal overall for various use cases. Do we think there is another way to move forward of being able to PDA or so forth. we need to have that figure out

NeilS: this discussion happened almost 2 years ago and I think we need to circle back and ask these questions to SR vendors

mhakkinen: may be it is time to create a poll to present the approaches to SR vendors if they would do any of them or all of them

janina: it doesnt need two be only SR. but we need to independent implementation

paulG: what is any changes need to be made in AX tree

NeilS: I dont know how hard it is for them to go back to original DOM but it is worth to ask this question to browser vendor

Paulg: it makes more sense to ask this questionf rom them than just sending a poll

PaulG: asking what is the accessible name in AX tree

NeilS: this is good point to know what is it mapped to

janina: I dont recall if we ever defined that

NeilS: I dont know what would it map pronunciation to

janina: updating the Ax tree would be helpful here

PaulG: we would have to have a discussion with browser vendors and AT vendors to understand creating new works

PaulG: role could be a role but it would be a new role like break role.. which will be new to AT, browser..

NeilS: attribute has to go somewhere

PaulG: anything is gotta go to axtree... it involves lot of mapping

PaulG: the alternative is that AT has to parse everything and for that reason they have lot to do

janina: we are syaing that we are just about AT and not path forward for PDAs because I dont see them looking at axtree or how they pronounce something. Am I wrong about that?

PaulG: no. they already processing SSML given to them by API

janina: do we have easy names for various direction we are pointing to? we should start with internal group such as WCIG. we have this proposal for WHATWG that they can accept. what do we call other ways to go back to this.

PaulG: we had single attr, multiple attr, and SSML native as first class object. and then each of those could be crossed with metrics with AT parsing for axtree solutions. we actually have six options. three ways for authors to add the info. two path forward for how it would process by speech tech

PaulG: either way AT is going to do heavy lift.

paulG: they probably are not trained to parse JSON so single attr is difficult compare to multiple attr because it is just DOM parsing

PaulG: SSML is native to those technologies. some of it has its prop. such as Amazon has a list. if we can get SSML parsing to the axtree by the user agent then I have pretty high hope that it could be a path forward. But then question who should we get involved

PaulG: if the AT vendors are fine with processing SSML for speech, and if it is all on axtree, experience will be different for AT vendors

NeilS: other issues could be related to speech engines because they all dont support SSML

NeilS: JAWS doesn't work with android

NeilS: ANdroid is using modern speech engines that support SSML. My main point was that you need to ship SSML to speech engines

PaulG: there may be voice packages

PaulG: looks like there are more decisions to make. DO we want to use the time between 9 and now to determine with if we have anyway to comeup with data driven approach

janina: that shoulds like a good simple question

PaulG: we give them choice between self parse or user agent parse

NeilS: sounds reasonable

janina: should we also ask why?

PaulG: one survey question and one already filled option

NeilS: are you planning to send survey or just an email

PaulG: this is the question of our strategy

NeilS: seems liek you are more likely to get informed response if we have a personal contacts

<IrfanAli> s/liel/like/

NeilS: I think we know people at vispero or NVDA. do we know decision makers at orca?

NeilS: at Apple is it James Craig? in that case we have reasonable contacts

mhakkinen: perhaps MS narrator and Peter Corn at Amazon

PaulG: I will send a DRAFT and you guys work on personal contacts and perhaps we have a plan for the next year

janina: at what point do we need to go back to WHATWG. should be engage them with this or after this poll?

PaulG: depends how this survey goes. we could use the leverage the survey asking them if they have preference for self parsing VS user agent parsing

other business

janina, re chartring. we may need to look into our deliverables. before July 31st 2023 we would like to update our charter

janina: we do not want to get stuck like last time. we have reasons to believe that we may be have charter in may/june, depending our deliverables

janina: our goal is not to be out of charter but we are not looking for any extension. we are going to look for the DRAFT in github

Minutes manually created (not a transcript), formatted by scribe.perl version 197 (Tue Nov 8 15:42:48 2022 UTC).


Succeeded: s/we thing/we think

Succeeded: s/need to/ need two

Succeeded: s/laternative/alternative

Succeeded: s/suppose/support

Failed: s/liel/like/

Maybe present: janina, mhakkinen, paulG

All speakers: janina, mhakkinen, NeilS, paulG

Active on IRC: IrfanAli, NeilS