Accessible Platform Architectures Working Group Teleconference

17 February 2021


janina, jasonjgw, joconnor, JPaton, SteveNoble

Meeting minutes

RAUR updates, if any.

janina: Item 14 - call participants and status - preserving data after the call

<joconnor> JOC: Is that a preservation of meta data thing?

janina: Knowing who was in the call

Judy: need to be cautious about capturing meeting chats unless participants know this in advance

<joconnor> JOC: So how is that an a11y thing?

janina: knowing the name of the speaker is very useful

Judy: also a common request from deaf participants

Smart agent/natural language dialogue system accessibility.

<joconnor> https://www.w3.org/WAI/APA/wiki/Voice_agent_user_requirements

Judy: Voice agent seems a bit restrictive

joconnor: The wiki sets up various definitions, and then drills down more to voice agents to limit scope

jasonjgw: If it doesn't have text input and output that can be an accessibility issue

<joconnor> JW: we need to look at Natural language processing and dialogue systems

jasonjgw: Some work in natural language has improved voice/text translation

judy: Concerned that the topic has "voice agent" in the header - it's disability restrictive

Judy: Smart agent or digital assistant would be more inclusive

janina: Consider web interface modality as a starting place - different paradigms - some have screen some do not - but here we have primarily a conversational approach

joconnor: Primarily these tools focus on voice

joconnor: Problem with the "big" umbrella is that is too large

JPaton: Makes sense to look at the whole digital assistant universe at some point, but makes sense to drill into more defined interfaces like voice assistants

janina: Maybe "conversational assistant?"

janina: If we can limit to certain knowledge domains we would get more meaningful results

Judy: use cases to consider such as someone who is deaf, quiet speech, etc.

Judy: the voice agent area may be very intriguing from a research perspective, but framing the discussion at a more cross-disability level at the start would be more helpful

jasonjgw: The term "conversational agent" may be more inclusive -- allows for text input/output and so forth

joconnor: We need a good definition to move forward

jasonjgw: Natural language includes a much larger focus - not dependent upon speech

janina: This is why we had VoiceXML spec, but it never seemed to be implemented

<Zakim> Judy, you wanted to comment on helping with the definition

<Judy> [JB: NLP is a decades-old subfield of linguistics, and we can take advantage of these kinds of established definitions to help us frame our note]

JPaton: Some systems include speech and text - not just speech

jasonjgw: Could write something to start?

joconnor: We need help on the definitions from others

Judy: Agree that we need definitions, but doubt it can be done very quickly - make take more time to develop solid underpinnings

jasonjgw: Can begin with writing the definition and then we can flesh out

jasonjgw: Action for next meeting to work on the definition assigned to Jason

XAUR: discussion prior to preparing for publication of wide review draft.

Minutes manually created (not a transcript), formatted by scribe.perl version 127 (Wed Dec 30 17:39:58 2020 UTC).


Succeeded: s/may seem/it's/

Succeeded: s/may be more/would be more/

Maybe present: Judy