AssistantJS
"develop ambitious platform-agnostic voice applications with ease"
About me
- Toni, 24 years old
- Working as conversational services architect @ Web Computing in Münster
- Masterthesis about platform-agnostic voice applications
- Maintainer of AssistantJS
Today's agenda
1. Basics of conversational applications
2. Hands on: Developing our own caring bot
Basics of
conversational applications
Participating components
Our Job
?
The assistant's backend
ASR = Automatic speech recognition
NLU = Natural language understanding
NRE = Named entity recognition
DM = Dialogmanager
?
NLU buzzwords
Utterance: Exemplary statements leading to intent
Intent: Intention of a user's voice query
Entity: Variables of a statement (e. g. "How do I get to the [train station]?")
Participating components
Intents and states
Developing a carebot
1) Generate new application
2) Integrate Alexa
3) Add first intents
4) Add response variations
5) Add general intents
6) Add second state
7) Add entity validation
8) Add google assistant
Additional features
- Platform-specific tests with Jasmine
- Strategy-based authentication
- Rich module-dependent logging
Conclusion
Our app's capabilities
- App starts the conversation with "How was your day?" and understands goodDayIntent and badDayIntent
- If the user has had a bad day, app asks for reason
- User is able to blame a mean colleague, app plots the revenge
- App detects if user did not mention the name of the mean colleague and follows up for it.
- App varies all given responses automatically and is multi-language-capable
- App works with Amazon Alexa and Google Assistant - on all devices
STAR BUTTON
(a.k.a. "the one to press")
AssistantJS repository
Wanna build the future of conversational interfaces?
WE'RE HIRING!
Will Google Assistant sound better anytime soon?
AssistantJS
By Antonius Ostermann
AssistantJS
A brief overview of AssistantJS, including a "hands on" tutorial to create a voice app using AssistantJS. (http://assistantjs.org)
