- From: Dirk Schnelle-Walka <dirk.schnelle@jvoicexml.org>
- Date: Thu, 19 Dec 2019 16:46:27 +0100 (CET)
- To: public-voiceinteraction@w3.org
Participants
- Bev Corwin
- Brian Susko
- Deborah Dahl
- Dirk Schnelle-Walka
general questions to the architecture
- core dialog
- provides interaction capabilities also if no dialogs are known
- dialog registry
- sort of a market place
- multiple providers may implement the ame intents
- Provider selection service would be in charge to selct the right one
- also consider what was previously selected
Dialog context
- who is in charge to resolve "it"?
- candidates are dialog manager or NLU of the IPA providers
- Dirk to add a "contextual knowledge" component to the figure
- context of the conversation must be kept outside the IPA Providers
- will work as long as the dialog is using a single providers
- unlikely to get this standardized
- is this covered in the current EMMA standard?
Dirk is planning to present the concpet draft at SpeechTek 2020
standardization possibilities
- Interface between Provider Selection Service and IPA Providers
- use EMMA JSON or any other flavor of it for input
- extensions might be needed on the input for the IPA Providers
- to return values or take actions
- do text first and other things later?
- Brian to extend the architecture for the interfaces
- Dirk to share the Visio in GitHub
- all other boxes might be subject to standardization, except IPA Providers
- specifically diaglog registry
- authentication
- possible standards exist
- not trying to reinvent the wheel
Hooks to IoT
- scritpting standard exist
- Debbie will look into this
Next meeting 16.01.
Received on Thursday, 19 December 2019 15:46:33 UTC