Systems and methods for proactively predicting user intents in personal agents

1. A system comprising:

a computing device configured to:

obtain user intent data identifying a desired action based on a first input by a user on a network-enabled tool;

obtain contextual data characterizing a user's interaction with the network-enabled tool;

determine at least one predicted future intent of the user on the network-enabled tool based on the user intent data and the contextual data using a predictive user intent model, wherein the at least one predicted future intent identifies at least one suggested action that is suggested to be performed by the user after the desired action;

cause the network-enabled tool to automatically complete the desired action and automatically present the at least one suggested action to the user on the network-enabled tool, before the user provides any additional input after the first input; and

update the predictive user intent model using updated data including the user intent data and the contextual data obtained by the system to replace synthetic data, wherein the system comprises a user intent prediction system that can continuously update the predictive user intent model, wherein the user intent data and the contextual data is collected during use of the system.

2. The system of claim 1 , wherein the network-enabled tool is an electronic personal agent or voice assistant.

3. The system of claim 1 , wherein the user intent data comprises user current query data and user conversation data from a current session with the network-enabled tool and the contextual data comprises user preferences or current status of the current session.

4. The system of claim 1 , wherein the user intent data and the contextual data are obtained in real-time during a common session by the user on the network-enabled tool.

5. The system of claim 1 , wherein the automatically presenting the at least one suggested action to the user comprises causing the network-enabled tool to display or audibly communicate the at least one suggested action as a selectable action of a plurality of selectable actions.

6. The system of claim 1 , wherein the determining the at least one predicted future intent comprises applying the user intent data and the contextual data to the predictive user intent model trained with the synthetic data.

7. The system of claim 6 , wherein the computing device is further configured to:

determine an evaluation metric that characterizes a performance of the predictive user intent model;

compare the determined evaluation metric to an archived evaluation metric of an archived predictive user intent model; and

replace the predictive user intent model with the archived predictive user intent model when the evaluation metric indicates that the predictive user intent model is less accurate than the archived predictive user intent model.

8. A method comprising:

obtaining user intent data identifying a desired action based on a first input by a user on a network-enabled tool;

obtaining contextual data characterizing a user's interaction with the network-enabled tool;

determining at least one predicted future intent of the user on the network-enabled tool based on the user intent data and the contextual data using a predictive user intent model, wherein the at least one predicted future intent identifies at least one suggested action that is suggested to be performed by the user after the desired action;

automatically completing the desired action and automatically presenting the at least one suggested action to the user on the network-enabled tool, before the user provides any additional input after the first input; and

updating the predictive user intent model using updated data including the user intent data and the contextual data obtained by the system to replace synthetic data, wherein the predictive user intent model can be continuously updated, and wherein the user intent data and the contextual data is collected during use of the system.

9. The method of claim 8 , wherein the network-enabled tool is an electronic personal agent or voice assistant.

10. The method of claim 8 , wherein the user intent data comprises user current query data and user conversation data from a current session with the network-enabled tool and the contextual data comprises user preferences or current status of the current session.

11. The method of claim 8 , wherein the user intent data and the contextual data are obtained in real-time during a common session by the user on the network-enabled tool.

12. The method of claim 8 , wherein the presenting the automatically at least one suggested action to the user comprises causing the network-enabled tool to display or audibly communicate the at least one suggested action as a selectable action of a plurality of selectable actions.

13. The method of claim 8 , wherein the determining the at least one predicted future intent comprises applying the user intent data and the contextual data to the predictive user intent model trained with the synthetic data.

14. The method of claim 13 , further comprising:

determining an evaluation metric that characterizes a performance of the predictive user intent data model;

comparing the determined evaluation metric to an archived evaluation metric of an archived predictive user intent model; and

replacing the predictive user intent model with the archived predictive user intent model when the evaluation metric indicates that the predictive user intent model is less accurate than the archived predictive user intent model.

15. A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by at least one processor, cause a device to perform operations comprising:

obtaining user intent data identifying a desired action based on a first input by a user on a network-enabled tool;

obtaining contextual data characterizing a user's interaction with the network-enabled tool;

determining at least one predicted future intent of the user on the network-enabled tool based on the user intent data and the contextual data using a predictive user intent model, wherein the at least one predicted future intent identifies at least one suggested action that is suggested to be performed by the user after the desired action;

automatically completing the desired action and automatically presenting the at least one suggested action to the user on the network-enabled tool, before the user provides any additional input after the first input; and

updating the predictive user intent model using updated data including the user intent data and the contextual data obtained by the system to replace synthetic data, wherein the predictive user intent model can be continuously updated, and wherein the user intent data and the contextual data is collected during use of the system.

16. The non-transitory computer readable medium of claim 15 , wherein the network-enabled tool is an electronic personal agent or voice assistant.

17. The non-transitory computer readable medium of claim 15 , wherein the user intent data comprises user current query data and user conversation data from a current session with the network-enabled tool and the contextual data comprises user preferences or current status of the current session.

18. The non-transitory computer readable medium of claim 15 , wherein the automatically presenting the at least one suggested action to the user comprises causing the network-enabled tool to display or audibly communicate the at least one suggested action as a selectable action of a plurality of selectable actions.

19. The non-transitory computer readable medium of claim 15 , wherein the predictive user intent model is trained with the synthetic data.

20. The non-transitory computer readable medium of claim 19 , having instructions stored thereon, wherein the instructions, when executed by the at least one processor, cause the device to perform further operations comprising:

determining an evaluation metric that characterizes a performance of the predictive user intent model;

comparing the determined evaluation metric to an archived evaluation metric of an archived predictive user intent model; and

replacing the predictive user intent model with the archived predictive user intent model when the evaluation metric indicates that the predictive user intent model is less accurate than the archived predictive user intent model.