Pronunciation Task Force Teleconference

10 Jul 2019


Sam_K, JF, irfan, paul_grenier, Dee, janina, Chris, Roy


<irfan> chair: irfan

<scribe> scribe: JF

New Membership

irfan: discussion about potential new members

JS: need to follow up on that - bad timing last week

Irfan: Have a new web dev from Maryland who is interested. Following our work already
... he has some potential use-cases

JS: makes sense to have him join. Need to connect him to Michael and move forward

Irfan: I have some other potential candidates - will forward that to janina

TPAC 2019

irfan: need to provide a list of groups we want to coordinate with
... noted some other discussion points on the ARIA WG mailng list related to SSML (etc.)

want to get agenda updated

JS: has irfan updated mark re: Leonie's comments of a few weeks ago?
... there is some good news re: getting SSML into HTML
... but wonders if we want AT to parse html and find the SSML doc, or do we need to to an AAM (accessibility API mapping)
... figure that is a Mark H question - do we want to approach the screen readers for their feedback?

irfan: remember this discussion and the people we need to chat with

MH: when discussed this with different companies, the idea of an AT vendor of parsing the SSML document out of the HTML document appeared to be a non-starter

JS: so the AAM may be the killer application

MH: now that we have allowed time to pass, we should perhaps query the various vendors and gauge their reaction to the different models we're looking at
... we have different authoring scenarios, so getting their feedback now would be useful

JS: do we have enough time before TPAC?

MH: yes. Do we do this verbally, or do we forward a more formal survey mechanism? have we done this before? how formal?

JS: we don't need to be *too* formal

MH: thinking we assemble an example suite of markup options

JF: suggest we have something of a formal paper-trail

MH: suggest that next step is to put together sample page(s) to socialize

irfan: should we be talking at TPAC about this too?

JS: yes, but how many will be in Japan?

MH: correct, will the right people be there? We might contemplate organizing something for the Wednesday

irfan: I believe that most of the people will be there

MH: we should get the survey out before TPAC

get some feedback prior to TPAC - may impact our discussion there

User Scenarios

irfan: any further updates?

SK: no, I am just settling back in after moving

at a point where can validate gathering more data, to re-enforce what we have

we have a body of knowledge already... in terms of their applicability

what i am able to contribute is to validate assumptions already

in terms of more content... unless there is more coming believe we've exhausted them already

reminder that they are scenarios, and not use-cases

perhaps by TPAC we can refine them to use-cases

irfan: saw them today. Discussed with Roy and we'll start adding them to the [..]

SK: yes, they are there waiting for more comments
... have been approached by other UX folks, so it is being socialized

when we have other docs ready, we will need to go through taht and test whether the docs clearly explain its purpose

Chris: when Roy does his pull, how do we then make more additons?

irfan: keep adding to git-hub, we can track there

JS: request to irfan - can you provide the git-hub URIs in the agendas

Use case document

irfan: thanks to Paul. Have looked at this, and have discussed this with Mark.

MH: in process of writing response to Paul
... issue is that Mac does not formally support SSML, and will likely not, as Apple has their own markup

PG: we can be held hostage by Apple, or we can ask the vendors to scrape the page and we move forward, and Mac can lag behind

MH: right, or we can do as we've done in the past and write up some JS to polyfill this

PG: ya, we'd need to do that.

irfan: worth trying to chat with james craig about this?

MH: I think the example of using the component markup is the best idea, as it is the AT that will need to grapple with this

PG: yes, AT is the less prepared for this right now
... I also submitted another issue re: webVTT. Not sure how that would work with SSML

<Chris> Must drop off. Thanks.

MH: have spoken with David Singer about this prior, and also Nigel from BBC

he was interested in using this when BBC text was read by TTS in the BBC style

irfan: wondering about Texthelp working with Apple?

MH: they are also using thier own TTS service, so they may be by-passing the Apple service

they are essentially injecting an SSML based code

irfan: need to look at their documentation

PG: Google's text API takes SSML already

send it as a web-request, and Google sends back streaming mp3s

right now there are no limits because there isn't a lot of usage, but it may emerge with a price model

MH: yes, it's pennies per utterance

PG: if you write your own intermediary, you may be able to by-pass this, as you are collecting these mp3 files

so you pay to store those files instead of contstantly waiting for real-time

but all of this may not be cross-platform, so this is likely a fall-back

PG: so, the other option is that if this is an AT option, then the AT can do as they best consider

benefit there is that it could support other voice-packs

so that if new ones emerge, they could swap-out

irfan: this is something interesting to see, to see how Google is serving this to the Mac speech API

do we know how they are doing this?

MH: all being done in the cloud'

PG: primary use-case for embedding SSML directly in the page - i.e. a news org can get names pronounced correctly
... especially in 'dramatic readings' it would add more depth (variances, speech rate and tone, etc.)

JS: wondering about the business model?

PG: like a Prime subscription, or the google subscription, etc.
... but when i looked at them, they are all using SSML variants

MH: so Apple is the only one that doesn't support SSML

irfan: wondering if it's a good time to migrate the documents

PG: the data code is ready, along with some of my comments

rather than having the best ideas swallowed up, perhaps we short list thme

so we should likely narrow it down to 2 or 3 options

JF: think we should first be broad before we narrow things down

irfan: now is a good time to work on the use-cases more, and think it is time to start getting this into a single document space


PG: notice that pull requests are being accepted, but not merged

Gap Analysis

MH: have done another update to document. REady for comments - have more to do but actively workin on this

irfan: we also discussed items from the CSS speech document - we need to look at that more closely and look to perhaps add some stuff

MH: yes, we need to discuss further


irfan: discussed with Roy earlier - we are on track, and still making progress

next week we will review these docs, and then if need be re-assess our timeline

expect that by end of July, we'll need to review timeline and adjust as required

JS: yes, looks like we will publish before TPAC

other business

irfan: anything?


trackbot, end meeting

Summary of Action Items

Summary of Resolutions

[End of minutes]

Minutes manually created (not a transcript), formatted by David Booth's scribe.perl version 1.154 (CVS log)
$Date: 2019/07/10 14:50:46 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.154  of Date: 2018/09/25 16:35:56  
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/ifran/irfan/
Succeeded: s/Apople/Apple/
Succeeded: s/erving/serving/
Succeeded: s/sop/so/
Succeeded: s/bing/being/
Succeeded: s/ifran/irfan/
Default Present: Sam_K, JF, irfan, paul_grenier, Dee, janina, Chris, Roy
Present: Sam_K JF irfan paul_grenier Dee janina Chris Roy
Found Scribe: JF
Inferring ScribeNick: JF
Found Date: 10 Jul 2019
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]