Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Build conversational interfaces for your customers

Build conversational interfaces for your customers

alking and listening is the most natural way to interact, we learn to do so since the very day we were born. In this session, I will show you how to build great conversational UI to delight your customers. We will cover the basic of speech recognition and natural language processing, explore the main programming interface and best practices to create engaging conversational interfaces. I will illustrate these concepts with Amazon Lex and Amazon Polly.

More Decks by Sébastien Stormacq - AWS Developer Advocate

Other Decks in Technology

Transcript

  1. © 2020, Amazon Web Services, Inc. or its Affiliates. Sébastien

    Stormacq Principal Developer Advocate, AWS @sebsto An Introduction to Conversational Interfaces
  2. © 2020, Amazon Web Services, Inc. or its Affiliates. 70s

    80s 90s 00s Present Graphical User Interface (GUI) Character Mode Voice User Interface (VUI) Mobile Web
  3. © 2020, Amazon Web Services, Inc. or its Affiliates. HUMAN

    ACCURACY 50% 55% 60% 62% 70% 95% ASR accuracy has dramatically increased in the last 4-5 years. This inflection point has created sustained momentum in consumer adoption of voice technology Source: MindMeld
  4. © 2020, Amazon Web Services, Inc. or its Affiliates. VOICE

    REPRESENTS THE NEXT MAJOR DISRUPTION IN COMPUTING
  5. © 2020, Amazon Web Services, Inc. or its Affiliates. “

    There’s no future that doesn’t have ambient computing or voice activation ” Mark Cuban, Entrepreneur
  6. © 2020, Amazon Web Services, Inc. or its Affiliates. How

    it works Request Audio Response Your Service Text to Speech Machine Learning Natural Language Understanding Speech Recognition Cards
  7. © 2020, Amazon Web Services, Inc. or its Affiliates. Wake

    word detection Beam forming Cards Your Service Text to Speech Machine Learning Natural Language Understanding Speech Recognition Request Response Signal Processing
  8. © 2020, Amazon Web Services, Inc. or its Affiliates. Speech

    to text Your Service Text to Speech Machine Learning Natural Language Understanding Speech Recognition Response Cards Utterances Intents
  9. © 2020, Amazon Web Services, Inc. or its Affiliates. ASR

    – Automatic Speech Recognition fo ̇ r tē tīmz Forty Times? 40x For Tea Times? For Tee Times?
  10. © 2020, Amazon Web Services, Inc. or its Affiliates. Utterances

    Into Intents Your Service Text to Speech Machine Learning Natural Language Understanding Speech Recognition Response Cards Utterances Intents
  11. © 2020, Amazon Web Services, Inc. or its Affiliates. Alexa,

    wake word S P E A K I N G W I T H A L E X A
  12. © 2020, Amazon Web Services, Inc. or its Affiliates. Alexa,

    ask restaurant finder for breakfast wake word Invocation name launch S P E A K I N G W I T H A L E X A utterance for a good breakfast place to find places to eat in the morning a spot that serves the most important meal of the day
  13. © 2020, Amazon Web Services, Inc. or its Affiliates. Utterances

    Intent Utterances and Intents RecomendationIntent to find places to eat in the morning a good breakfast place where to eat breakfast
  14. © 2020, Amazon Web Services, Inc. or its Affiliates. Alexa,

    open restaurant finder Intents and Slots Recommend a restaurant within four miles slot utterance RecommendationIntent {distance: “4”} slot value intent
  15. © 2020, Amazon Web Services, Inc. or its Affiliates. Tell

    me about Italian food within four miles slot RecommendationIntent {distance: “4”} slot value intent slot {cuisine: “Italian food”} Intents and Slots
  16. © 2020, Amazon Web Services, Inc. or its Affiliates. Requests

    and Responses Your Service Text to Speech Machine Learning Natural Language Understanding Speech Recognition Response Cards Audio Intents
  17. © 2020, Amazon Web Services, Inc. or its Affiliates. Speech

    Output Your Service Text to Speech SSML, streaming audio Machine Learning Natural Language Understanding Speech Recognition Response Cards Audio Intents
  18. © 2020, Amazon Web Services, Inc. or its Affiliates. SSML

    • Speech Synthesis Markup Language • TTS (Text-to-Speech)
  19. © 2020, Amazon Web Services, Inc. or its Affiliates. SSML

    • Pauses well, okay well <break time="3s"/> okay
  20. © 2020, Amazon Web Services, Inc. or its Affiliates. SSML

    • Pronunciation Hints <say-as interpret-as="spell-out"> Hello </say-as>
  21. © 2020, Amazon Web Services, Inc. or its Affiliates. SSML

    • Speechcons <say-as interpret-as="interjection"> Hurray! </say-as>
  22. © 2020, Amazon Web Services, Inc. or its Affiliates. The

    Brain of Alexa in your Apps Amazon Lex
  23. © 2020, Amazon Web Services, Inc. or its Affiliates. Designed

    for builders; efficient tools to build conversations; scales automatically Enterprise ready: Connect to enterprise systems via SaaS connectors Build once and deploy to multiple platforms Text and speech language understanding: Powered by the same technology as Alexa Amazon Lex: Conversational engine
  24. © 2020, Amazon Web Services, Inc. or its Affiliates. Should

    I develop for Amazon Alexa or Amazon Lex? * Images and video on Echo Show, Echo Spot, and other devices with a screen Amazon Alexa Amazon Lex Apps Skills Bots Medium Voice and Displays* DIY (rendering) Out of the Box Built-in + Skills Nothing Distribution Skill Store DIY (your app or device) Devices Alexa Enabled Devices (Amazon & 3P) Yours (web, mobile, hardware) Reach Alexa Customers Your Customers * Images and video on Echo Show, Echo Spot, and other devices with a screen
  25. © 2020, Amazon Web Services, Inc. or its Affiliates. Utterances

    Utterances are spoken or typed phrases that invoke your intent BookHotel Intents Intents perform an action in response to natural language user input Slots Slots are input data required to fulfill the intent Fulfillment Fulfillment mechanism for your intent . Amazon Lex
  26. © 2020, Amazon Web Services, Inc. or its Affiliates. Check

    in 4/19/2017 City New York City I would like to book a hotel Sure, which city will you be traveling to? New York City What date do you want to check in? April 19th Slot elicitation
  27. © 2020, Amazon Web Services, Inc. or its Affiliates. Simple

    Declarative Model Slots Prompts Build multi-turn conversations Easy setup in console Sure, which city will you be traveling to? New York City What date do you want to check in? Tomorrow … And how many nights is this for? Dialog management City Check-In Date Check-Out Date Which city will you be traveling to? What date do you want to check in? How many nights is this for? I would like to book a hotel
  28. © 2020, Amazon Web Services, Inc. or its Affiliates. Would

    you prefer to stay downtown this time as well? Yes What date do you want to check in? April 19th Personalize conversation based on user preferences Validate user input and re-prompt as necessary Lambda function Customize conversations Sorry, no availability. Would a different location work? I would like to book a hotel
  29. © 2020, Amazon Web Services, Inc. or its Affiliates. Sorry,

    can you please repeat that? I am having trouble understanding. Can you please say that again? Clarifies by requesting that user repeat Hang-up phrase to end the conversation Error handling Sorry, I am not able to assist you at this time. I want to …. {garbled} ….. Uses a different prompt every time
  30. © 2020, Amazon Web Services, Inc. or its Affiliates. Compact

    • Formatted for messaging platforms • Multiple cards supported • Preview capability • Test in console Rich message formatting Sure. What type of car are you looking for? I would like to rent a car
  31. © 2020, Amazon Web Services, Inc. or its Affiliates. AWS

    Lambda integration Return to client User input is parsed to derive intents and slot values; output is returned to client for further processing Intents and slots are passed to the AWS Lambda function for business logic implementation Fulfillment
  32. © 2020, Amazon Web Services, Inc. or its Affiliates. Start

    with people, not computers Source: BBC, Research & Development, Prototyping for Voice: Methodology – bit.ly/bbc-vui (last accessed August 2019)
  33. © 2020, Amazon Web Services, Inc. or its Affiliates. Pay

    attention to details Plan for escapes Test Be brief #1 #10 #9 #8 #7 #6 Be coherent #5 Offer multiple choices #4 Re-engage conversation #3 Stay natural #2 Be gender neutral Teach customers Tips for great conversations
  34. © 2020, Amazon Web Services, Inc. or its Affiliates. Thank

    you! Sébastien Stormacq Principal Developer Advocate, AWS @sebsto