Introduction
Curious about how speech technology really works, but afraid to ask? Evaluating the claims of conversational AI to build something that works well? Have you built a voice skill or action and wondered if people will like talking to it or how to improve it? Working with a team that’s experienced in their fields, but not in voice? Or wanting to understand why voice-experienced people insist on doing certain things certain ways? Maybe you’re just tired of talking to conversational devices that aren’t responding the way you’d like them to!
Wherever you’re at, welcome to the world of voice-first development! And thanks for picking up this sampler of three chapters from our book, Voice-First Development, parts of which are available now through MEAP (Manning Early Access Program)!
Today’s explosion of voice solutions is exciting. Widely available platforms allow practically anyone to build conversational voice solutions. That's outstanding. But availability doesn’t guarantee success or hordes of happy users. Most resources today focus on platform-specific development. A few discuss design topics. And others look at a single slice of creating voice solutions, like content creation or monetization. What’s been missing is an in-depth treatment of start-to-end voice deployment based on real world data and experiences. One where design and development are treated as intertwined equals. Silos have no place in successful conversational voice development. We want to help you understand the How and Why of great voice solutions so that you can sift through hype and promises and create voice solutions that users love with the voice technology available today.
Part 1 of our book gives you a solid foundation of both the technology behind today’s voice-first solutions and how people produce and understand speech. Chapter 1, included here, introduces the core concepts of conversational voice while investigating the reality behind some claims and beliefs of voice development. You’ll learn about the interconnected modules of voice architecture and the purpose of each phase of voice development.
We’ve also included chapters 7 and 8, which introduce development thinking, code and design techniques relevant to a specific voice-first topic, just as each chapter does in Part 2 of our book. Throughout, we intentionally blend design, development, and product owner concerns because we want to encourage everyone to have a grasp of the whole pie, not just their own slice. Chapter 7 focuses on how to handle vague, unclear or ambiguous user requests, and how and when to disambiguate. Disambiguation is a core concept in voice solutions, because you can’t limit what people say but your solution needs clarity before it can fulfill the user’s request. Chapter 8 shows you why voice solutions need to reflect the appropriate level of certainty, how and when to ask users for confirmation, and what can go wrong if the wrong assumptions are made.
Many examples used in these chapters come from two made up voice solutions: a restaurant finder and a medical procedure prep assistant. Both are heavily based on real ones and draw on our experiences in creating voice solutions across platforms and devices. You’ll learn how to recognize and avoid common pitfalls on each topic, as well as address them when they do happen. You’ll find lots of actionable steps and success tips, stories from the field, and concrete design tips and code samples to help you learn and understand best practices for creating voice solutions. Our code samples mainly use Google’s Dialogflow and the Actions on Google framework for convenience, but you can easily apply the solutions to other frameworks.
Above all, we share with you what we’ve learned during decades of building and analyzing voice solutions. Some solutions were great, and we’re proud of those. Others, well not as much. You’ll be hearing about those too—they’re often the more interesting ones.
So come along and learn from our mistakes and our successes – and build great voice interactions for us all to use!
—Ann Thymé-Gobbel and Charles Jankowski