When a user initiates a conversation, how the system responds to the user’s request is determined by a set of rules called a dialog model.
The system matches the request to a known intent (disambiguation), and collects required information (slot filling) to execute the users intent. The model consists of five parts:
Determine the User’s Intent
The initial user request may or may not include all the information needed by the system to understand the intent. Based on the request, the system will match:
The system identifies one, and only one, possible user intent with a high level of certainty. If it succeeds in doing that, it will move on to the next part of the dialog model.
Disambiguate, if Necessary
Even though the system has been able to identify a single intent, there may be parts of the user’s utterance that require further clarification.
Gather Information, if Necessary
Once the system has determined the correct intent, additional information (slot filling) may be required from the user to complete the corresponding action or request.
The system will request that the user provides the missing information.
Review, if Necessary
For creating and updating items, the system will prompt the user to review and confirm any actions before the transaction is executed.
Once the system has gathered all the information it needs, it will complete the request or action. Based on the type of intent, this can result in the following: