Scalable Endpoint-Dependent Natural Language Understanding

• US 20170263255A1
• Filed: 03/10/2016
• Published: 09/14/2017
• Est. Priority Date: 03/10/2016
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Jeong, Minwoo, Sarikaya, Ruhi
• Granted Patent: US 10,229,687 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/33 Querying
• CPC Class Codes: G06F 16/90332 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 40/35 Discourse or dialogue representation
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications

Abstract

A computer-implemented technique is described for processing a linguistic item (e.g., a query) in an efficient and scalable manner. The technique interprets the linguistic item using a language understanding (LU) system in a manner that is based on a particular endpoint mechanism from which the linguistic item originated. The LU system may include an endpoint-independent subsystem, an endpoint-dependent subsystem, and a ranking component. The endpoint-independent subsystem interprets the linguistic item in a manner that is independent of the particular endpoint mechanism. The endpoint-dependent subsystem interprets the linguistic item in a manner that is dependent on the particular endpoint mechanism. The ranking component generates final interpretation results based on intermediate results generated by the endpoint-independent subsystem and the endpoint-dependent subsystem, e.g., by identifying the most likely interpretation of the linguistic item.

Claims

1. A system, implemented by one or more computing devices, for processing a linguistic item, the linguistic item corresponding to information expressed in a natural language, the system comprising:

an interface component configured to receive a linguistic item from a particular device-implemented endpoint mechanism, in response to interaction by a user with an input device associated with the particular endpoint mechanism,

the particular endpoint mechanism corresponding to one of a set of different types of endpoint mechanisms;

an endpoint-independent subsystem for interpreting the linguistic item in a manner that is independent of the particular endpoint mechanism from which the linguistic item originated, the endpoint-independent subsystem generating first intermediate results;

an endpoint-dependent subsystem for interpreting the linguistic item in a manner that is dependent on the particular endpoint mechanism from which the linguistic item originated, the endpoint-dependent subsystem generating second intermediate results; and

a ranking component configured to generate final interpretation results based on the first intermediate results and the second intermediate results,

the final interpretation results representing an interpretation of a meaning of the input linguistic item.

2. The system of claim 1, wherein the particular endpoint mechanism corresponds to a particular type of user computing device through which the user enters the linguistic item.

3. The system of claim 1, wherein the particular endpoint mechanism corresponds to a program component through which the user enters the linguistic item, as implemented by a computing device.

4. The system of claim 3, wherein the program component corresponds to a browser component or a messaging flow component.

5. The system of claim 1,

wherein the endpoint-independent subsystem includes one or more endpoint-independent interpreter components,

wherein the endpoint-dependent subsystem includes one or more endpoint-dependent interpreter components,

wherein each endpoint-independent interpreter component is configured to interpret linguistic items directed to a particular domain, in a manner that is independent of endpoint mechanisms from which the linguistic items originated, and

wherein each endpoint-dependent interpreter component is configured to interpret linguistic items directed to a particular domain, in a manner that is dependent on endpoint mechanisms from which the linguistic items originated.

6. The system of claim 5, wherein one particular endpoint-independent or endpoint-dependent interpreter component includes a domain determination component configured to determine whether the linguistic item that has been received corresponds to a particular domain.

7. The system of claim 5, wherein one particular endpoint-independent or endpoint-dependent interpreter component includes an intent determination component configured to determine an intent associated with the linguistic item.

8. The system of claim 5, wherein one particular endpoint-independent or endpoint-dependent interpreter component includes a slot value determination component configured to extract one or more slot values from the linguistic item.

9. The system of claim 5,

wherein said one or more endpoint-dependent interpreter components associated with the endpoint-dependent subsystem generate one or more instances of interpreter results, and

wherein the endpoint-dependent subsystem further includes a post-processing component configured to filter the interpreter results based on endpoint information that identifies the particular endpoint mechanism that has been used to provide the linguistic item.

10. The system of claim 1,

wherein the ranking component identifies a candidate interpretation result (CIR) item considered to be a most likely interpretation of the linguistic item, selected from among the first intermediate results and the second intermediate results, and

wherein the ranking component identifies the CIR item based at least on endpoint information that identifies the particular endpoint mechanism that has been used to provide the linguistic item.

11. The system of claim 1, wherein the endpoint-independent subsystem and/or the endpoint-dependent subsystem and/or the ranking component is implemented, at least in part, using one or more machine-learned models.

12. The system of claim 1, wherein the endpoint-independent subsystem and/or the endpoint-dependent subsystem and/or the ranking component is implemented, at least in part, using one or more rule-application engines.

13. The system of claim 1, wherein an action-taking component is configured to perform a computer-implemented action in response to the final interpretation results.

14. A method, implemented by one or more computing devices, for processing a linguistic item, the linguistic item corresponding to information expressed in a natural language, the method comprising:

receiving a linguistic item from at a device-implemented endpoint mechanism, in response to interaction by a user with an input device associated with the particular endpoint mechanism,

the particular endpoint mechanism corresponding to one of a set of different types of endpoint mechanisms;

processing the linguistic item by a language understanding (LU) system to generate final interpretation results,

the LU system including;

an endpoint-independent subsystem for interpreting the linguistic item in a manner that is independent of the particular endpoint mechanism from which the linguistic item originated, to generate first intermediate results; and

an endpoint-dependent subsystem for interpreting the linguistic item in a manner that is dependent on the particular endpoint mechanism, to generate second intermediate results; and

performing a computer-implemented action in response to the final interpretation results.

15. The method of claim 14, wherein the particular endpoint mechanism corresponds to a particular type of user computing device and/or a program component through which the user enters the linguistic item.

16. The method of claim 14, wherein the LU system further includes a ranking component configured to generate final interpretation results based on the first intermediate results and the second intermediate results.

17. The method of claim 14,

wherein the endpoint-independent subsystem includes one or more endpoint-independent interpreter components,

wherein the endpoint-dependent subsystem includes one or more endpoint-dependent interpreter components,

wherein each endpoint-independent interpreter component is configured to interpret linguistic items directed to a particular domain, in a manner that is independent of endpoint mechanisms from which the linguistic items originated, and

wherein each endpoint-dependent interpreter component is configured to interpret linguistic items directed to a particular domain, in a manner that is dependent on endpoint mechanisms from which the linguistic items originated.

18. The method of claim 17, further including accommodating a new endpoint mechanism or application without affecting at least one existing interpreter component of the LU system.

19. A computer-readable storage medium for storing computer-readable instructions, the computer-readable instructions, when executed by one or more processor devices, providing a language understanding (LU) system that comprises:

an endpoint-independent subsystem configured to interpret a linguistic item in a manner that is independent of a particular device-implemented endpoint mechanism from which the linguistic item has been received, the endpoint-independent subsystem generating first intermediate results, the linguistic item corresponding to information expressed in a natural language,

the endpoint-independent subsystem including one or more endpoint-independent interpreter components,

each endpoint-independent interpreter component being configured to interpret linguistic items directed to a particular domain, in a manner that is independent of the particular endpoint mechanisms from which the linguistic items originated;

an endpoint-dependent subsystem configured to interpret the linguistic item in a manner that is dependent on the particular endpoint mechanism, the endpoint-dependent subsystem generating second intermediate results,

the endpoint-dependent subsystem including one or more endpoint-dependent interpreter components,

each endpoint-dependent interpreter component being configured to interpret linguistic items directed to a particular domain, in a manner that is dependent on the particular endpoint mechanisms from which the linguistic items originated; and

a ranking component configured to identify a candidate interpretation result (CIR) item that is considered to be a most likely interpretation of the linguistic item, selected from among the first intermediate results and the second intermediate results.

20. The computer-readable storage medium of claim 19, wherein the particular endpoint mechanism corresponds to a particular type of user computing device and/or a program component through which the user enters the linguistic item.