The document discusses the development and architecture of voice user interfaces (VUIs), highlighting the transformation of an existing app into an advanced architecture that supports VUI applications. It covers key concepts such as natural language processing and domain routing while providing implementation examples in node.js. Additionally, it emphasizes the importance of a solid architecture for effective VUI design and shares lessons learned from the process.

1. Control your Voice
like a Bene Gesserit
Jorge D. Ortiz Fuentes
@jdortiz

2. A Canonical
Examples
Production

3. @jdortiz
Agenda
★ Context
★ Voice User Interfaces 101
★ Advanced Architecture Brief
★ VUI + Adv. Architecture
★ Lessons Learned

4. Warning

5. Sorry!

6. Put all your devices
in non listening
mode

7. Some
Context

8. Living like a 3rd
Stage Guild Navigator

9. @jdortiz
The Request
★ Existing App
★ Refactored to an Advanced Architecture
★ Wanted to explore VUI
★ Prototype with minimum cost

10. Voice User
Interfaces
101

11. Welcome to the
Future

12. @jdortiz
Maybe not as good as we
wished (YET)
★ “Sorry. I don’t know that one” “Sorry about
that. I’m still learning”
★ English, German, French, Japanese (Voice
interaction)
★ Private actions

13. The Contenders
Alexa Google Home SiriCortana

14. – Bene Genesserit
“I must not fear. Fear is the mind-killer.
Fear is the little-death that brings total
obliteration. I will face my fear.”

15. The Guts
Domain 1
Domain 2
Domain 3
Domain n
Routing
Natural
Language
Processing
Automatic
Speech
Recognition
Text to
Speech
Voice ➡
Voice
Text
Domain &
Intent &
Parameters
Text

16. Domain nDomain n
Extend the System
JSON ➡
JSON
HTTPS
Parse
Request
Do Your
Stuff
Build
Response

17. @jdortiz
Development
★ Node JS library (Alexa Skills & Google
Actions)
★ Also Alexa SDK for Java(/Kotlin)
★ Several unofficial ones

18. @jdortiz
var routes = Routes()
routes.add(method: .post, uri: “/") { request, response in
defer {
response.completed()
}
if let postString = request.postBodyString,
let data = postString.data(using:.utf8) {
let reply = AlexaServiceResponse(version: "1.0",
response: AlexaResponse(outputSpeech:
OutputSpeech(text: "Hello. I am your Genie. How can I help
you?",
type: .plaintext)))
let encoder = JSONEncoder()
let responseData = try! encoder.encode(reply)
let responseString = String(bytes: responseData, encoding: .utf8) !?? ""
response.setHeader(.contentType, value: "application/json")
response.setBody(string: responseString)
} else {
response.badRequest(message: "No request data found.")
}
}

19. @jdortiz
var routes = Routes()
routes.add(method: .post, uri: “/") { request, response in
defer {
response.completed()
}
if let postString = request.postBodyString,
let data = postString.data(using:.utf8) {
let reply = GAV1ServiceResponse(speech:
"Hello. I am your Genie. How can I help you?",
data:
gaServiceRequest.originalRequest.data)
let encoder = JSONEncoder()
let responseData = try! encoder.encode(reply)
let responseString = String(bytes: responseData, encoding: .utf8) !??
""
response.setHeader(.contentType, value: "application/json")
response.setBody(string: responseString)
} else {
response.badRequest(message: "No request data found.")
}
}

20. OK
But…
Why do I need a
good architecture?

21. Advanced
Architecture

22. @jdortiz
Persistance FW
View
Network
LocationFW
Presenter
Entity Gateway
Clean Architecture
Interactor
Entity

23. Advanced Architecture:
Mobile
App
Delegate /
Application
View (VC)
/ Activity
Presenter Use Case
Entity
Gateway
Connector
Tap on
phone button
User wants
to talk to
contact
Start call
if…
Get
contact
details

24. @jdortiz
Injecting Dependencies
View
Presenter
UseCaseFactory
Entity
Gateway
Connector

25. VUI +
Advanced
Architecture

26. Kwisatz Haderach

27. Advanced Architecture
VUI
Server
UI Presenter Use Case
Entity
Gateway
Connector
Reuse
Domain
Logic
Intent
StartCall
User wants
to talk to
contact
Start call
if…
Get
contact
details

28. Lessons
Learned

29. – Dune, Frank Herbert
“Muad'Dib learned rapidly because his first
training was in how to learn. And the first
lesson of all was the basic trust that he
could learn… Muad'Dib knew that
every experience carries its lesson.”

30. The not-so-relevant
Technical Stuff

31. @jdortiz
Implementation Details
★ Rebuild your Model (Alexa Skills)
★ Lots of corner cases in Codable
★ Documentation of the JSON requests /
responses could be better

32. @jdortiz
var routes = Routes()
routes.add(method: .post,
uri: “/alexa/“,
handler: alexaRequestHandler)
routes.add(method: .post,
uri: “/gaction/“,
handler: gactionRequestHandler)
let server = HTTPServer()
server.addRoutes(routes)
server.serverPort = UInt16(httpPort)
do {
try server.start()
} catch PerfectError.networkError(let err, let msg) {
LogFile.debug("Network error thrown: (err) (msg)")
}

33. @jdortiz
func gactionRequestHander(request: HTTPRequest, response: HTTPResponse) {
defer {
response.completed()
}
LogFile.info("!>> Google Action Service Request received !>>")
if let postString = request.postBodyString,
let data = postString.data(using:.utf8) {
do {
let decoder = JSONDecoder()
let gaRequest = try decoder.decode(GAV1ServiceRequest.self, from: data)
let reply = try gaction.process(serviceRequest: gaRequest)
let encoder = JSONEncoder()
let responseData = try! encoder.encode(reply)
let responseString = String(bytes: responseData, encoding: .utf8) !?? ""
LogFile.debug("Response: (responseString)")
response.appendBody(string: responseString)
response.setHeader(.contentType, value: "application/json")
} catch let error as GActionError {
response.append(error: error)
LogFile.error("Google Action error: (error)”)
} catch let error {
response.badRequest(message: "Request data is wrong.")
LogFile.error("Error decoding request: (error)")
}
} else {
response.badRequest(message: "No request data found.")
LogFile.error("No request data found.")
}
}

34. @jdortiz
let skill = AlexaSkill(id: appId,
welcomeMessage: "Welcome to Family Chores.
How can I help you?",
logger: GenericLogger())
skill.register(controller: RecordTaskAlexaUIController(),
forIntent: "RecordTask")
let connector = PersonReportAlexaConnector()
skill.register(controller: PersonReportAlexaController(),
forIntent: "PersonReport")

35. @jdortiz
protocol AlexaUIController {
func process(serviceRequest: ASServiceRequest) throws !-> ASServiceResponse
}
class AlexaSkill { !//…
func process(serviceRequest: ASServiceRequest) throws !-> ASServiceResponse {
guard serviceRequest.safetyCheck(id: id) else {
throw SkillError.wrongSkillId
}
switch serviceRequest.request.type {
case .launch:
return serviceRequest.reply(message: welcomeMessage)
case let ASRequestType.intent(data: intentData):
if let controller = controllerMap[intentData.name] {
return try controller.process(serviceRequest: serviceRequest)
} else {
throw SkillError.undefinedIntent
}
case .sessionEnd:
return serviceRequest.end()
}
}
}

36. Identity & Flow

37. @jdortiz
Flow
★ Don’t make your users talk their way
through
★ Straight from A -> B

38. But wait!
There is concurrency

39. @jdortiz
skill.register(controller: PersonReportAlexaController(),
forIntent: "PersonReport")

40. @jdortiz
skill.register(connector = PersonReportAlexaConnector(),
forIntent: "PersonReport")

41. Fast Prototyping

42. @jdortiz
Quality Time
★ Avoid duplication to address both platforms
★ Ignore the DB
★ Reuse and extend use cases

43. Recap

44. @jdortiz
Key Takeaways
★ Having (/Going to) an Advanced
Architecture is key
★ Pros & Cons of own API interface
★ Easy Prototyping
★ Speak what you wish /Use what you wish

45. Bedankt!

46. Thank
You!

47. @jdortiz