Using Intents to Analyze and Personalize a User's Dialog Experience with a Virtual Personal Assistant

• US 20140310001A1
• Filed: 07/30/2013
• Published: 10/16/2014
• Est. Priority Date: 04/16/2013
• Status: Active Grant

• Current Assignee: SRI International, Inc.
• Original Assignee: SRI International, Inc.
• Inventors: Mark, William S., Kalns, Edgar T., Ayan, Necip Fazil
• Granted Patent: US 9,875,494 B2
• Time in Patent Office: Days
• US Class Current: 704/270.1
• CPC Class Codes: G06Q 30/06 Buying, selling or leasing transactions
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems

Abstract

A virtual personal assistant (VPA) application analyzes intents to, among other things, enhance or personalize a user'"'"'s dialog experience with the VPA application. A set of intents, or multiple sets of intents, are maintained over the course of one or more user-specific dialog sessions with the VPA. Inferences may be derived from the set or sets of intents and incorporated into a current or future dialog session between the VPA and a user of the VPA application. In some embodiments, the inferences are only made available through the systemic understanding of natural language discourse by the VPA.

Claims

1. A method for personalizing a user experience with a computerized virtual personal assistant (“

VPA”

) application, the VPA application enabling user-directed natural language dialog between a user and the VPA application, the method comprising, with a computing system;

maintaining a chain of intents, the chain of intents comprising a plurality of temporally-spaced intents, each of the intents representing a structured interpretation of natural language dialog expressed by a user of the VPA application during a conversational natural language dialog session involving the user and the VPA application;

recognizing a pattern of intents in the chain of intents;

deriving an inference from the pattern of intents, the inference comprising a conclusion about the user or about the dialog session between the user and the VPA application that is not otherwise available to the VPA application; and

applying the inference to a subsequent conversational natural language dialog session involving the VPA application and the user.

2. The method of claim 1, comprising determining a goal likely intended to be accomplished by the user during the dialog session, inferring from the pattern of intents that the goal has been fulfilled, and incorporating the inference that the goal has been fulfilled into output generated by the VPA application during the subsequent dialog session.

3. The method of claim 1, comprising determining a goal likely intended to be accomplished by the user during the dialog session, inferring from the pattern of intents that the goal has not been fulfilled, and incorporating the inference that the goal has not been fulfilled into output generated by the VPA application during the subsequent dialog session.

4. The method of claim 3, comprising, during the subsequent dialog session, presenting natural language output relating to the goal.

5. The method of claim 1, comprising inferring from the pattern of intents a preference, opinion, or characteristic of the user, and adapting output generated by the VPA application during the subsequent dialog session to incorporate the user'"'"'s preference, opinion, or characteristic.

6. The method of claim 1, comprising adjusting an interpretation of natural language dialog expressed by the user during the subsequent dialog session based on the inference derived from the pattern of intents.

7. The method of claim 1, wherein each of the intents in the chain of intents comprises an input intent, the method comprises associating one or more output intents with each of the input intents, and each of the output intents relates to output generated by the VPA application in response to one of the input intents.

8. The method of claim 1, wherein the chain of intents comprises a plurality of temporally-spaced input intents and output intents, and each of the output intents relates to output generated by the VPA application in response to one of the input intents.

9. The method of claim 1, wherein the chain of intents comprises intents representing interpretations of natural language dialog expressed by a user of the VPA application during a plurality of different conversational natural language dialog sessions involving the user and the VPA application, and the method comprises recognizing the pattern of intents across the plurality of different dialog sessions.

10. The method of claim 1, wherein the method is embodied as computer-executable instructions stored in one or more machine accessible storage media.

11. A virtual personal assistant (“

VPA”

) computer application embodied in one or more machine-accessible storage media, the VPA computer application comprising;

a user interface through which a user-directed conversational natural language dialog session may be conducted between a computing device and a user; and

a VPA engine to conduct the dialog session and initiate the providing of a service or information to the user in response to conversational natural language dialog supplied by the user during the dialog session, the VPA engine configured to;

maintain a dialog context, the dialog context comprising a plurality of temporally-spaced intents, each of the intents representing at least a partial interpretation of natural language dialog expressed by the user during the dialog session;

determine a current intent, the current intent representing a partial interpretation of natural language dialog expressed by the user subsequent to the natural language dialog represented by the intents in the dialog context;

identify an intent in the dialog context having a data relationship with the current intent;

combine the identified intent with the current intent in accordance with the data relationship; and

generate system output based on the combined intent.

12. The VPA computer application of claim 11, wherein the VPA engine is configured to determine an intent mapping, the intent mapping defines the data relationship between the current intent and the identified intent, and the VPA engine is configured to combine the identified intent with the current intent in accordance with the intent mapping.

13. The VPA computer application of claim 12, wherein the VPA engine is configured to determine a relationship between a parameter of the current intent and a parameter of the identified intent based on the intent mapping, and add a data value associated with the parameter of the identified intent with the current intent.

14. The VPA computer application of claim 12, wherein the intent mapping identifies a field of the current intent as similar to a field of the identified intent.

15. The VPA computer application of claim 11, wherein the current intent represents a partial interpretation of natural language dialog expressed by the user subsequent to an intervening instance of user-supplied conversational natural language dialog, and the intervening instance is temporally subsequent to the natural language dialog represented by the identified intent.

16. The VPA computer application of claim 11, wherein each intent in the dialog context has a status, and the VPA engine is configured to select the identified intent based on the status of the identified intent.

17. The VPA computer application of claim 11, wherein the VPA engine is configured to search the dialog context for intents that have a specified status, parameter, or data value, and select the identified intent from a set of intents that have the specified status, parameter, or data value.

18. The VPA computer application of claim 11, wherein the VPA engine is configured to select the identified intent based on an inference, and the inference is derived from a pattern of intents associated with a plurality of different conversational natural language dialog sessions involving the user and the VPA computer application.

19. The VPA computer application of claim 11, wherein each of the intents including the combined intent has a status of complete or incomplete, and if the status of the combined intent is complete, the VPA engine is configured to determine whether another intent in the dialog context has a status of incomplete, and if another intent in the dialog context has a status of incomplete, the VPA engine is configured to add the combined intent to the dialog context and use the dialog context to interpret the other incomplete intent.

20. A computing device comprising machine accessible storage media and a processor, wherein at least a portion of the VPA computer application of claim 11 is stored in the machine accessible storage media and the VPA computer application is executable by the processor.

21. An intent analyzer to analyze a user experience with a computerized virtual personal assistant (“

VPA”

) application, the VPA application enabling user-directed natural language dialog between a user and the VPA application, the intent analyzer embodied in one or more machine accessible storage media and configured to;

maintain a set of intents, one or more of the intents representing a structured interpretation of natural language dialog expressed by a user of the VPA application during a conversational natural language dialog session involving the user and the VPA application;

identify a pattern of intents in the set of intents;

derive an inference about the dialog session from the pattern of intents; and

apply the inference to another conversational natural language dialog session involving the VPA application.

22. The intent analyzer of claim 21, wherein the set of intents comprises an intent representing an interpretation of natural language dialog expressed by another user of the VPA application.

23. The intent analyzer of claim 21, configured to incorporate the inference into one or more other conversational natural language dialog sessions involving the VPA application and one or more other users.

24. The intent analyzer of claim 21, comprising deriving the inference from an interpretation of non-dialog input received by the computer system.

25. The intent analyzer of claim 21, wherein the intent analyzer is embodied as machine readable instructions executable by a computing device.

26. The intent analyzer of claim 21, wherein the intent analyzer is embodied as one or more modules of the VPA application.