Christmas came early this year for Alexa developers, because Amazon has finally granted one of our fondest wishes! This month a default, Fallback Intent for Alexa skills was added to the interaction model.
The Fallback Intent Is Added Automatically
We’ve always had to anticipate how users will interact with our skills, but it’s impossible to predict every possible thing a user might say and that makes it impossible to catch every possible error. Many a dev knows the agony of a cert fail with a note attached saying your skill doesn’t gracefully recover from user speech that doesn’t map to your skill’s intents.
Next time you head to the Developer Console to create a new Alexa skill, you’ll find a new intent has been added to the list of default, automatically included intents from Amazon: the AMAZON.FallbackIntent. From Justin Jeffress, writing for Amazon’s Alexa developer blog:
The new AMAZON.FallbackIntent in the Alexa Skills Kit (ASK) built-in library helps you handle unexpected utterances, or when a customer says something that doesn’t map to any intents in your skill. As you design your skill, you identify and define a set of utterances and intents to capture the many possible variations that a customer can speak. When your skill is active, Alexa will try to match all customer utterances to one of the skill’s intents. There may be instances when the customer request is not appropriate for your skill. If Alexa maps one of these requests to your intents, it may provide an unexpected response.
We refer to these types of requests as “out-of-domain requests.” The new AMAZON.FallbackIntent lets you respond gracefully to an out-of-domain request to your Alexa skill. You can provide a specific handler for this intent in your Alexa skill to provide additional instructions or sample utterances on what the skill does and reorient your customers. For example, the trip planner skill illustrated below has mapped utterances (within the skill), to the skill’s intent. However, if a customer speaks an utterance that is out-of-domain (outside of the skill), unexpected, or does not map to any of your skill intents, then the skill can invoke the AMAZON.FallbackIntent to redirect the customer to utterances that are mapped to an intent or provide additional information.
Additionally, you can use the new Intent Request History API that offers an aggregated and anonymized view of customer’s intent requests to evolve your skill’s intents or fine-tune the out-of-domain model as necessary.
Click here to read the full post on Amazon’s developer blog, which includes instructions for implementing the new Fallback intent and relevant resource and documentation links.