Method and system for facilitating a guided dialog between a user and a conversational agent

• US 10,403,273 B2
• Filed: 09/09/2016
• Issued: 09/03/2019
• Est. Priority Date: 09/09/2016
• Status: Active Grant

• Current Assignee: Verizon Patent and Licensing Incorporated (Verizon Communications Inc.)
• Original Assignee: Oath Inc. (Verizon Communications Inc.)
• Inventors: Lee, Sungjin, Stent, Amanda
• Primary Examiner(s): Le, Thuykhanh
• Application Number:
  US15/260,596
• Publication Number:
  US 20180075847A1
• Time in Patent Office: 1,089 Days
• Field of Search: None
• CPC Class Codes: G06F 16/24522 Translation of natural language queries to structured queries
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• 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 2015/0635 updating or merging of old and new templates; Mean values; Weighting
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command

Abstract

The present teaching relates to facilitating a guided dialog with a user. In one example, an input utterance is obtained from the user. One or more task sets are estimated based on the input utterance. Each of the one or more task sets includes a plurality of tasks estimated to be requested by the user via the input utterance and is associated with a confidence score computed based on statistics with respect to the plurality of tasks in the task set. At least one of the one or more task sets is selected based on their respective confidence scores. A response is generated based on the tasks in the selected at least one task set. The response is provided to the user.

Claims

1. A method, implemented on a machine having at least one processor, storage, and a communication platform connected to a network for facilitating a guided dialog with a user, the method comprising:

obtaining an input utterance from the user; and

guiding a dialog with the user by;

determining one or more task sets estimated based on the input utterance, wherein each of the one or more task sets includes a plurality of tasks estimated to be requested by the user via the input utterance, and is associated with a confidence score computed based on statistics with respect to the plurality of tasks in the task set, and a time interval between the input utterance and a last utterance by the user, wherein the confidence score is inversely proportional to the time interval;

selecting at least one of the one or more task sets based on their respective confidence scores;

generating a response based on the tasks in the selected at least one task set; and

providing the response to the user.

2. The method of claim 1, wherein determining one or more task sets comprises:

determining a plurality of dialog act items based on a semantic analysis of the input utterance;

assigning each of the plurality of dialog act items to a task frame candidate to generate a first set of task frame candidates, wherein at least one of the plurality of dialog act items is shared by two or more task frame candidates;

determining a first task set that includes a plurality of tasks corresponding to the first set of task frame candidates;

determining a task slot value for each of the plurality of dialog act items based on the assignment with respect to the first set of task frame candidates; and

calculating a first confidence score associated with the first task set.

3. The method of claim 2, wherein determining one or more task sets further comprises:

reassigning at least one of the plurality of dialog act items to a different task frame candidate to generate a second set of task frame candidates;

determining a second task set that includes a plurality of tasks corresponding to the second set of task frame candidates;

determining a task slot value for each of the plurality of dialog act items based on the reassignment with respect to the second set of task frame candidates; and

calculating a second confidence score associated with the second task set.

4. The method of claim 2, wherein determining a task slot value for each of the plurality of dialog act items includes:

incorporating one or more Boolean logics; and

determining one or more complex task slot values for at least one of the plurality of dialog act items, based on the one or more Boolean logics.

5. The method of claim 2, wherein the first confidence score is calculated based on at least one of the following statistics:

the number of task frames in the first set of task frame candidates;

the number of dialog act items assigned to each task frame candidate;

the average number of dialog act items per active task frames in the first set of task frame candidates;

the entropy of dialog act item distribution across the active task frames; and

the number of active task frames with only one dialog act item.

6. The method of claim 1, further comprising:

extracting one or more task lineages of the user from a task state database, wherein each of the one or more task lineages includes a sequence of tasks previously estimated to requested by the user;

generating one or more new task lineages based on the selected at least one task set;

determining coherent features between the extracted one or more task lineages and the one or more new task lineages;

estimating a belief score for each new task lineage based on the coherent features and the confidence scores of the tasks in the selected at least one task set; and

extending each of the extracted one or more task lineages of the user based on the estimated belief scores.

7. The method of claim 6, further comprising:

determining relevant context information based on the extracted one or more task lineages and the selected at least one task set;

determining a context window representing a time period within which the relevant context information can be fetched;

retrieving the relevant context information from a database based on the context window; and

updating at least one task slot value of the sequence of tasks in the extracted one or more task lineages based on the retrieved relevant context information.

8. The method of claim 1, wherein the statistics with respect to the plurality of tasks in the task set include a number of dialogue act items assigned to a task frame candidate.

9. The method of claim 1, wherein at least one of a plurality of dialog act items is shared by two or more task frame candidates.

10. A system, having at least one processor, storage, and a communication platform connected to a network for facilitating a guided dialog with a user, comprising:

a semantic analyzer implemented by the at least one processor and configured for obtaining an input utterance from the user;

a task frame parser implemented by the at least one processor and configured for determining one or more task sets estimated based on the input utterance, wherein each of the one or more task sets includes a plurality of tasks estimated to be requested by the user via the input utterance and is associated with a confidence score computed based on statistics with respect to the plurality of tasks in the task set, and a time interval between the input utterance and a last utterance by the user, wherein the confidence score is inversely proportional to the time interval;

a dialog state updater implemented by the at least one processor and configured for selecting at least one of the one or more task sets based on their respective confidence scores; and

a response generator implemented by the at least one processor and configured for generating a response based on the tasks in the selected at least one task set, and providing the response to the user.

11. The system of claim 10, wherein determining one or more task sets comprises:

determining a plurality of dialog act items based on a semantic analysis of the input utterance;

assigning each of the plurality of dialog act items to a task frame candidate to generate a first set of task frame candidates, wherein at least one of the plurality of dialog act items is shared by two or more task frame candidates;

determining a first task set that includes a plurality of tasks corresponding to the first set of task frame candidates;

determining a task slot value for each of the plurality of dialog act items based on the assignment with respect to the first set of task frame candidates; and

calculating a first confidence score associated with the first task set.

12. The system of claim 11, wherein determining one or more task sets further comprises:

reassigning at least one of the plurality of dialog act items to a different task frame candidate to generate a second set of task frame candidates;

determining a second task set that includes a plurality of tasks corresponding to the second set of task frame candidates;

determining a task slot value for each of the plurality of dialog act items based on the reassignment with respect to the second set of task frame candidates; and

calculating a second confidence score associated with the second task set.

13. The system of claim 11, wherein determining a task slot value for each of the plurality of dialog act items includes:

incorporating one or more Boolean logics; and

determining one or more complex task slot values for at least one of the plurality of dialog act items, based on the one or more Boolean logics.

14. The system of claim 11, wherein the first confidence score is calculated based on at least one of the following statistics:

the number of task frames in the first set of task frame candidates;

the number of dialog act items assigned to each task frame candidate;

the average number of dialog act items per active task frames in the first set of task frame candidates;

the entropy of dialog act item distribution across the active task frames; and

the number of active task frames with only one dialog act item.

15. The system of claim 10, wherein the dialog state updater further comprises:

a task lineage extractor implemented by the at least one processor and configured for extracting one or more task lineages of the user from a task state database, wherein each of the one or more task lineages includes a sequence of tasks previously estimated to requested by the user;

a slot state updater implemented by the at least one processor and configured for generating one or more new task lineages based on the selected at least one task set;

a coherent feature determiner implemented by the at least one processor and configured for determining coherent features between the extracted one or more task lineages and the one or more new task lineages;

a belief estimate determiner implemented by the at least one processor and configured for estimating a belief score for each new task lineage based on the coherent features and the confidence scores of the tasks in the selected at least one task set; and

a task lineage extender implemented by the at least one processor and configured for extending each of the extracted one or more task lineages of the user based on the estimated belief scores.

16. The system of claim 15, further comprising:

a relevant context determiner implemented by the at least one processor and configured for determining relevant context information based on the extracted one or more task lineages and the selected at least one task set;

a context window determiner implemented by the at least one processor and configured for determining a context window representing a time period within which the relevant context information can be fetched; and

a relevant context retriever implemented by the at least one processor and configured for retrieving the relevant context information from a database based on the context window, wherein the slot state updater is further configured for updating at least one task slot value of the sequence of tasks in the extracted one or more task lineages based on the retrieved relevant context information.

17. A non-transitory machine-readable medium having information recorded thereon for facilitating a guided dialog with a user, wherein the information, when read by the machine, causes the machine to perform the following:

obtaining an input utterance from the user; and

guiding a dialog with the user by;

determining one or more task sets estimated based on the input utterance, wherein each of the one or more task sets includes a plurality of tasks estimated to be requested by the user via the input utterance, and is associated with a confidence score computed based on statistics with respect to the plurality of tasks in the task set, and a time interval between the input utterance and a last utterance by the user, wherein the confidence score is inversely proportional to the time interval;

selecting at least one of the one or more task sets based on their respective confidence scores;

generating a response based on the tasks in the selected at least one task set; and

providing the response to the user.

18. The medium of claim 17, wherein determining one or more task sets comprises:

determining a plurality of dialog act items based on a semantic analysis of the input utterance;

assigning each of the plurality of dialog act items to a task frame candidate to generate a first set of task frame candidates, wherein at least one of the plurality of dialog act items is shared by two or more task frame candidates;

determining a first task set that includes a plurality of tasks corresponding to the first set of task frame candidates;

determining a task slot value for each of the plurality of dialog act items based on the assignment with respect to the first set of task frame candidates; and

calculating a first confidence score associated with the first task set.

19. The medium of claim 18, wherein determining one or more task sets further comprises:

reassigning at least one of the plurality of dialog act items to a different task frame candidate to generate a second set of task frame candidates;

determining a second task set that includes a plurality of tasks corresponding to the second set of task frame candidates;

determining a task slot value for each of the plurality of dialog act items based on the reassignment with respect to the second set of task frame candidates; and

calculating a second confidence score associated with the second task set.

20. The medium of claim 18, wherein determining a task slot value for each of the plurality of dialog act items includes:

incorporating one or more Boolean logics; and

determining one or more complex task slot values for at least one of the plurality of dialog act items, based on the one or more Boolean logics.

21. The medium of claim 18, wherein the first confidence score is calculated based on at least one of the following statistics:

the number of task frames in the first set of task frame candidates;

the number of dialog act items assigned to each task frame candidate;

the average number of dialog act items per active task frames in the first set of task frame candidates;

the entropy of dialog act item distribution across the active task frames; and

the number of active task frames with only one dialog act item.

22. The medium of claim 17, further comprising:

extracting one or more task lineages of the user from a task state database, wherein each of the one or more task lineages includes a sequence of tasks previously estimated to requested by the user;

generating one or more new task lineages based on the selected at least one task set;

determining coherent features between the extracted one or more task lineages and the one or more new task lineages;

estimating a belief score for each new task lineage based on the coherent features and the confidence scores of the tasks in the selected at least one task set; and

extending each of the extracted one or more task lineages of the user based on the estimated belief scores.

23. The medium of claim 22, further comprising:

determining relevant context information based on the extracted one or more task lineages and the selected at least one task set;

determining a context window representing a time period within which the relevant context information can be fetched;

retrieving the relevant context information from a database based on the context window; and

updating at least one task slot value of the sequence of tasks in the extracted one or more task lineages based on the retrieved relevant context information.