Web Speech:
Past, Present,
Future and Opportunity
Interesting historical note
There's history of us trying to make
machines that 'speak' or 'listen' going
back to at least the 1700s
Bell Labs demonstrated digital voice
synthesis at the World's Fair in 1939
If you find this interesting, I wrote a whole
piece on The History of Speech
Who remembers this movie?
1983
Me in 1983:
Me...
... I was easily distracted.
The Web: Early 90s - the web really brought me back to computers
<object ID="AgentControl" width="0" height="0" CLASSID="clsid:D45FD31B-5C6E-11D1-9EC1-00C04FD7081F" CODEBASE="http://server/path/msagent.exe#VERSION=2,0,0,0">
</object>
1997!!!
Me in 1997:
WebSpeech
The path to the current state of affairs...
1998 CSS2 Aural Stylesheets
March 1999
AT&T Corporation, IBM, Lucent, and Motorola formed the VoiceXML Forum
Handed over to W3C in 2000
AskJeves, AT&T, Avaya, BT, Canon, Cisco, France Telecon, General Magic, Hitachi, HP, IBM, isSound, Intel, Locus Dialogue, Lucent, Microsoft, Mitre, Motorola, Nokia, Nortel, Nuance, Phillips, PipeBeach, Speech Works, Sun, Telecon Italia, TellMe.com, and Unisys
Mayyyybbbeeee?
Like HTML, but for... wait... I'm confused.
So many XMLs
For the next decade...
2010: Speech XG Community Group
let's sidestep this mess...
Competing draft proposals from:
- Google,
- Microsoft,
- Mozilla
- Voxeo
Stir....
~ kinda a spec
Next stop:
Working Group
LOL jk
2012: Another Community Group!
~"Here's what we're shipping"
Buuuuuttt....
The Web: Present Day
Almost :(
The Web: Present Day
- Parts of it work everywhere (almost, better than many 'standards' in history)
- No RECs here, or even earlier. Not standards tracked
- The actual de-facto standard isn't written down
- Lots of partial successes
- Lots of interest from developers!
- No-one is prioritizing this
I wrote some posts: Web Speech (series)
- Intro: The History and State of Speech
- Part I: Greetings, Professor Falken
- Part II: You Don't Say
- Part III: Listen up
- Part IV: Thoughts on Voice Recognition
- It is historically weird (like MathML)
- Here's how it works (and doesn't)
- It needs resolution and attention to resolve its weirdness and architecture.. But nobody is doing it
- The world is quite a bit different too
- I have some thoughts about how to resolve this
We should ultimately have a well explained, layered architecture with escape hatches
-
A promise based API
-
Feature detection
-
A sensible voice model for TTS, like fonts.
-
Constructors/A replaceable service provider
Lots of service APIs
Even at the OS level!
Lots of discussion, but no driver.
So what's holding it up?
It lacks a driver - especially one without an interest in both ends of it
A few proposals.
Basic support for Linux would be great for WebKit in general :(
Opportunity and advantage Igalia?
Voice notifications?
Hands free assistance? (keep that screen clean)
"Play Frozen for the Kids in the back"
"Turn on/off/up the AC/Heat"
Voice notifications?
"You're due for an oil change soon"
WPE could have advantage with Speech.
Igalia could position as a strong driver at fixing historical problems...
Lots of potential funding strategies here potentially.
Thanks.
Web Speech APIs now
By Brian Kardell
