Chapter 13. Managing Failure
The quality of a voicebot is judged not only by how the voicebot manages its interactions with users when everything is going smoothly, but also (and perhaps more importantly) by how it navigates situations when things don’t proceed as expected.
In this chapter, we discuss recovery strategies that, if systematically observed, will markedly improve a voicebot’s performance and, therefore, the satisfaction of its users.
Types of Failure
First, let’s describe the four main types of failure that a user may encounter while engaging with a voicebot.
No-Input
A no-input failure is triggered when the voicebot is unable to detect sound from the user.
No-Match
A no-match failure is triggered when the voicebot is unable to interpret what the user said. In other words, the voicebot heard the user say something that the voicebot did not expect them to say.
Misrecognition
A misrecognition failure occurs when the voicebot thinks the user said something that is different from what the user actually said. Unlike the no-input and the no-match failures, misrecognition is a failure that the voicebot is not aware of. Two situations then present themselves: (1) The user is able to quickly detect that the voicebot misrecognized what the user said, or (2) The user does not detect that the voicebot misrecognized what the user said.
System Failure
A system failure occurs when something goes wrong at the application layer. For instance, a data source is not available. ...
Get The Elements of Voice First Style now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.