What kind of successful speech applications are out there. TellMe, Asterisk.
Speech interfaces not representing the low end. Letting people self service. TellMe is priced up too high. If Google wants to do something they should offer a low cost per minute service, pre-paid. SIP based for example.
Voice application gadgets, drag and drop the same way that the google homepage is. For instance voice search of NYC restaurants.
Rich brought up the point of initiation time. Push to talk for voice services. Better to call a friend. (But Google is your friend!). Asynchronous. Voice is cool, but better to not happen over the phone. Different transport. TellMe CEO has a call open all the time, like a conf call to a session on all the time. Wildfire? I haven’t seen that, could be worth checking out.
goog-411 uses VoiceXML.
Noise is a big problem, not a big problem. There doesn’t seem to be an agreement about that.
Mention of multimodal, voice and web, simultaneous.
Amount of voice presence huge, sms big, data not really there. Even with all the installs that google has gotten for its apps, the numbers aren’t significant.
what you need – VUI (voice UI, VoiceXML), Grammars (top movies, neighborhoods), community.
IMS (IP Multimedia System)? not there might never be there. complicated and ambitious. No real deployments yet.
What is the equivalent of text advertising in voice applications? Sponsored listings?
translation service. call something and say “ask this guy where the nearest restroom is in chinese”, and then pass the phone to him.
Private acoustic models, especially for people with accents.
machine learning, be able to correct the system and have it learn from that.
pick up stress level, emotion, feelings. So that you can pick up some additional context. That’s one already that acts as a lie detector.