Designing for speech
summary
Voice interfaces already hit scale — predictions claim 50% of searches by voice by 2020, 40% of adults use voice daily, Google Assistant on ~400M devices and Cortana ~145M monthly users, Echo Dot sells tens of millions. Practical guidance covers writing sample dialogues, grouping utterances into intents, using confidence threshold rules (explicit confirm between ~45–80%, auto-act above ~80%), choosing visual confirmation when ambiguity is likely, handling constrained versus conversational modes, parsing M-best recognition lists, and avoiding real-world failures like unwanted purchases, wrong wiki answers, or unlocking doors by accident.
AI-written summary · about
chapters (23)
- 0:00Intro
- 0:41How many of you have ever talked to Cortana
- 2:37Virtual assistants
- 2:55Words per minute
- 4:29Future predictions
- 5:36Cortana
- 5:56Alexa
- 6:27Google Assistant
- 7:31Future Tech
- 8:06Coding After Work
- 8:42Questions
- 13:17Best at Statement
- 16:13Sample Dialogues
- 18:04Voice User Interface Flow
- 20:01Visuals
- 23:13Example
- 24:58Confirmation
- 27:55Confidence
- 29:07Implicit Confirmation
- 31:19Nonspeech
- 35:26Conversational
- 40:16Discoverability
- 45:00Examples
related talks
59mDesigning for speech
45mKeynote: Conversations with Machines
52mDesigning digital assistants for those who need it most
21mHow Brands Must Leverage AI & Voice to Stay Relevant & Accessible
32mParty like it's 1970: Conversational interfaces are back (into your Android app)
01h 04mBuilding Expertise Into Generated Conversations
49mDesigning voice interfaces
56mAutomating Voice Assistants and Chatbots