Goal-oriented dialog system
A dialog manager receives text data corresponding to a dialog with a user. Entities represented in the text data are identified. Context data relating to the dialog is maintained, which may include prior dialog, prior API calls, user profile information, or other data. Using the text data and the context data, an N-best list of one or more dialog models is selected to process the text data. After processing the text data, the outputs of the N-best models are ranked and a top-scoring output is selected. The top-scoring output may be an API call and/or an audio prompt.
1. A computer-implemented method comprising:
receiving first input data corresponding to a first request of a dialog, the first input data comprising natural language;
determining the natural language includes an entity;
determining dialog history data representing the dialog;
determining first dialog focus data including:
the first input data,
an indicator corresponding to the entity, and
the dialog history data;
processing the first dialog focus data to determine:
a first score representing a first relationship between the first dialog focus data and a first trained model corresponding to a first category of functions, and
a second score representing a second relationship between the first dialog focus data and a second trained model corresponding to a second category of functions;
sending, based at least in part on the first score, the first dialog focus data to the first trained model;
sending, based at least in part on the second score, the first dialog focus data to the second trained model;
receiving, from the first trained model, first response data corresponding to a first potential response to the first request;
receiving, from the second trained model, second response data corresponding to a second potential response to the first request;
determining a third score representing a third relationship between the first input data and the first response data;
determining a fourth score representing a fourth relationship between the first input data and the second response data; and
performing, based at least in part on the third score and the fourth score, a first action associated with the first response data.
2. The computer-implemented method of claim 1 , further comprising:
sending the first dialog focus data to a third trained model corresponding to a third category of functions; and
receiving, from the third trained model, third response data corresponding to an application programming interface (API) call.
3. The computer-implemented method of claim 1 , further comprising:
determining a user account corresponding to the first input data;
determining user preference data corresponding to the user account;
determining that the user preference data indicates a negative preference of a system associated with the second trained model; and
determining the fourth score based at least in part on the negative preference.
4. A computer-implemented method comprising:
receiving first input data corresponding to a first command of a dialog, the first input data comprising natural language;
determining that the natural language includes an entity;
determining dialog history data representing the dialog;
determining first dialog focus data including:
the first input data,
a first indicator corresponding to the entity, and
the dialog history data;
processing the first dialog focus data to determine a first relationship between the first dialog focus data and a first model corresponding to a first category of functions;
sending, based at least in part on the first relationship, the first dialog focus data to the first model;
receiving, from the first model, first response data corresponding to a first potential response to the first command;
processing the first dialog focus data to determine a second relationship between the first response data and a first goal represented in the first input data;
determining, based at least in part on the second relationship, that a first action associated with the first response data is to be performed; and
performing the first action.
5. The computer-implemented method of claim 4 , further comprising:
determining second dialog focus data including the first dialog focus data and a second indicator corresponding to the first action.
6. The computer-implemented method of claim 4 , further comprising:
processing the first dialog focus data to determine a third relationship between the first dialog focus data and a second model corresponding to a second category of functions;
sending the first dialog focus data to the second model;
receiving, from the second model, second response data corresponding to the first command;
processing the first dialog focus data to determine a fourth relationship between the second response data and the first goal represented in the first input data; and
determining the first action further based at least in part on the fourth relationship.
7. The computer-implemented method of claim 4 , further comprising:
receiving second input data corresponding to a second command of the dialog;
determining second dialog focus data including at least the first dialog focus data and the second input data;
sending the second dialog focus data to a second model; and
receiving, from the second model, second response data corresponding to the second command.
8. The computer-implemented method of claim 4 , further comprising:
receiving first model data corresponding to the first model and an application programming interface (API);
determining that the first model data corresponds to the entity; and
determining the first relationship based at least in part on the first model data corresponding to the entity.
9. The computer-implemented method of claim 4 , further comprising:
determining second dialog focus data including at least the first dialog focus data and the first response data;
receiving second input data corresponding to a second command of the dialog;
sending the second dialog focus data to a second model; and
receiving, from the second model, second response data corresponding to the second command.
10. The computer-implemented method of claim 4 , further comprising:
determining a first candidate domain corresponding to the first input data;
determining a second candidate domain corresponding to the first input data; and
determining that the first input data corresponds to the entity based at least in part on processing the first input data using:
a first entity recognizer corresponding to the first candidate domain, and
a second entity recognizer corresponding to the second candidate domain.
11. The computer-implemented method of claim 4 , further comprising:
determining a user account corresponding to the first input data;
determining user preference data corresponding to the user account; and
determining the first dialog focus data to further include the user preference data.
12. A system comprising:
at least one processor; and
at least one memory including instructions that, when executed by the at least one processor, cause the system to:
receive first input data corresponding to a first command of a dialog, the first input data comprising natural language;
determine that the natural language includes an entity;
determine dialog history data representing the dialog;
determine first dialog focus data including:
the first input data,
a first indicator corresponding to the entity, and
the dialog history data;
process the first dialog focus data to determine a first relationship between the first dialog focus data and a first model corresponding to a first category of functions;
send, based at least in part on the first relationship, the first dialog focus data to a first model;
receive, from the first model, first response data corresponding to a first potential response to the first command;
process the first dialog focus data to determine a second relationship between the first response data and a first goal represented in the first input data;
determine, based at least in part on the second relationship, that a first action associated with the first response data is to be performed; and
perform the first action.
13. The system of claim 12 , wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the system to:
determine second dialog focus data including the first dialog focus data and a second indicator corresponding to the first action.
14. The system of claim 12 , wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the system to:
process the first dialog focus data to determine a third relationship between the first dialog focus data and a second model corresponding to a second category of functions;
send the first dialog focus data to the second model;
receive, from the second model, second response data corresponding to the first command;
process the first dialog focus data to determine a fourth relationship between the second response data and the first a goal represented in the first input data; and
determine the first action further based at least in part on the fourth relationship.
15. The system of claim 12 , wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the system to:
receive second input data corresponding to a second command of the dialog;
determine second dialog focus data including at least the first dialog focus data and the second input data;
send the second dialog focus data to a second model; and
receive, from the second model, second response data corresponding to the second command.
16. The system of claim 12 , wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the system to:
receive first model data corresponding to the first model and an application programming interface (API);
determine that the first model data corresponds to the entity; and
determine the first relationship based at least in part on the first model data corresponding to the entity.
17. The system of claim 12 , wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the system to:
determine second dialog focus data including at least the first dialog focus data and the first response data;
receive second input data corresponding to a second command of the dialog;
send the second dialog focus data to a second model; and
receive, from the second model, second response data corresponding to the second command.
18. The system of claim 12 , wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the system to:
determine a first candidate domain corresponding to the first input data;
determine a second candidate domain corresponding to the first input data; and
determine that the first input data corresponds to the entity based at least in part on processing the first input data using:
a first entity recognizer corresponding to the first candidate domain, and
a second entity recognizer corresponding to the second candidate domain.
19. The system of claim 12 , wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the system to:
determine a user account corresponding to the first input data;
determine user preference data corresponding to the user account; and
determine the first dialog focus data to further include the user preference data.
20. The system of claim 12 , wherein the first input data is automatic speech recognition (ASR) output data.