We are expanding support for FallbackIntent (previously available in different locations and languages) to Hindi (hi-IN).
If you are an Alexa Skills developer, you probably already know that the more effort you make to provide a comprehensive set of utterances in your skills, the better the results will be in understanding customer requests.
Customers often ask for requests in unusual ways. We refer to these types of requests as "requests outside the domain", that is, requests not present in the interaction model. The AMAZON.FallbackIntent allows Alexa skills to respond to unexpected requests at the time a client says something that is not mapped to any other intent with high assertiveness.
The addition of FallbackIntent to the interaction model of a skill offers benefits to both developers and users. FallbackIntent helps better target "out of domain" requests and allows the skill to provide a more relevant response to the user.
Skill developers using the Intent Request History API will now see "out-of-domain" requests in a demonimated category FallbackIntent, which will help to easily separate what should be covered in the skill interaction model. In this way, they will be able to explore new features, adjusting the voice model to the needs of users.
The following is an example of how the Fallback intent works:
User: Alexa, अंतरिक्ष facts शुरू करो
Alexa: अंतरिक्ष facts मैं आपका स्वागत हैं, आप क्या करना चाहेंगे?
User: मुझे शेर की आवाज़ सुनाओ (This utterance is not linked to any of the intents of the अंतरिक्ष facts skill.)
Since Alexa does not feel confident to associate this statement with any of the intents, it sends the skill to AMAZON.FallbackIntent.
Alexa: Sorry, अंतरिक्ष facts skill आपको इस मामले में मदद नहीं कर सकता, पर आपको अंतरिक्ष के दिलचस्प facts दे सकता हैं. क्या आप एक अंतरिक्ष fact जानना चाहेंगे? (The answer includes the name of the skill and offers help to reorient the user.)
When added to a skill, AMAZON.FallbackIntent automatically evaluates existing intentions and samples using machine learning, and creates a model for interacting with "out-of-domain" requests for the skill.
AMAZON.FallbackIntent can also be enhanced with additional phrases, allowing developers to adjust the interaction model if they wish. FallbackIntent is triggered when a customer request has a more reliable hypothesis mapping for AMAZON.FallbackIntent. In other words, the user's initial speech cannot be safely matched to any of the skill's other intentions. The skill backend receives an AMAZON.FallbackIntent request and can respond with helpful instructions or guidance.
The "out of domain" model for AMAZON.FallbackIntent is automatically generated from the skill's voice interface and a sample database from various existing sources. This model is generated again whenever an update to the voice interface occurs. So, it evolves with your skill as you improve your interaction model.
You will need to provide a handler on your backend that can give the customer an appropriate response when triggering FallbackIntent. FallbackIntentHandler is usually a prompt to inform the user that they have requested something that the skill cannot do, followed by guidance on the functionality of the skill. Your prompt should also remind the user that they are interacting with your skill. Here is an example using the Node.js SDK:
Implement AMAZON.FallbackIntent in your skill today to respond to unexpected requests from users more easily. Read the technical documentation for more information.
For more information, check out these resources: