Dynamically evolving cognitive architecture system based on training by third-party developers

• US 9,594,542 B2
• Filed: 08/18/2014
• Issued: 03/14/2017
• Est. Priority Date: 06/20/2013
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Viv Labs, Inc. (Samsung Electronics Co. Ltd.)
• Inventors: Gabel, Mark, Brigham, Christopher, Cheyer, Adam
• Primary Examiner(s): BROPHY, MATTHEW J
• Application Number:
  US14/461,876
• Publication Number:
  US 20140380286A1
• Time in Patent Office: 939 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/40 Processing or translation of natural language natural language analysis G06F40/20; semantic analysis G06F40/30
• CPC Class Codes: G06F 8/20 Software design
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition

Abstract

A dynamically evolving cognitive architecture system based on training by third-party developers is described. A system forms an intent based on a user input, and creates a plan based on the intent. The plan includes a first action object that transforms a first concept object associated with the intent into a second concept object and also includes a second action object that transforms the second concept object into a third concept object associated with a goal of the intent. The first action object and the second action object are selected from multiple action objects. The system executes the plan, and outputs a value associated with the third concept object.

Claims

1. A system for a dynamically evolving cognitive architecture based on training by third party developers, the system comprising:

one or more processors; and

a non-transitory computer readable medium storing a plurality of instructions, which

when executed, cause the one or more processors to;

receive a span of natural language annotated with an annotating object from a third-party developer, the annotating object comprising one of an input action object, an output action object, an input concept object, an intermediate concept object, another intermediate concept object and an output concept object;

match a first object and a second object with a user input, the first object comprising one of the input action object, the output action object, the input concept object, the intermediate concept object, the other intermediate concept object and the output concept object, the second object comprising another one of the input action object, the output action object, the input concept object, the intermediate concept object, the other intermediate concept object and the output concept object;

form an intent based on the user input;

create a plan based on the intent and the annotated span of natural language, the plan comprising the input action object selected from a plurality of action objects that transforms the input concept object associated with the intent into the intermediate concept object, and the output action object selected from the plurality of action objects that transforms the other intermediate concept object into the output concept object associated with a goal of the intent, the other intermediate concept object comprising one of a same object as the intermediate concept object and a different object from the intermediate concept object;

execute the plan; and

output a value associated with the output concept object associated with the user input.

2. The system of claim 1, wherein creating the plan is based on a confidence score associated with the annotated span of natural language.

3. The system of claim 1, wherein the plurality of instructions, when executed, will further cause the one or more processors to receive another span of natural language annotated with another annotating object from another third-party developer, wherein creating the plan is based on the other annotated span of natural language, and wherein the other annotating object comprises a different one of the input action object, the output action object, the input concept object, the intermediate concept object, the other intermediate concept object, and the output concept object.

4. The system of claim 1, wherein the plurality of instructions, when executed, will further cause the one or more processors to output a comparison based on a confidence score associated with the span of natural language annotated by the annotating object and a confidence score associated with the span of natural language without annotation by the annotating object, wherein the comparison indicates an amount of training completion relative to the span of natural language annotated with the object from the third-party developer.

5. The system of claim 1, wherein the plurality of instructions, when executed, will further cause the one or more processors to:

create a second plan based on the intent, wherein the second plan differs from the first plan by at least one concept object and by at least one action object; and

select the plan for execution based on at least one of an action object cost, an action object quality, a number of planned action objects, and a confidence score associated with the annotated span of natural language.

6. The system of claim 1, wherein the plurality of instructions, when executed, will further cause the one or more processors to output a prompt, based on the plan, which requests an additional user input, wherein executing the plan is based on the additional user input.

7. A computer-implemented method for a dynamically evolving cognitive architecture system based on training by third-party developers, the method comprising:

receiving a span of natural language annotated with an annotating object from a third-party developer, the annotating object comprising one of an input action object, an output action object, an input concept object, an intermediate concept object, another intermediate concept object and an output concept object;

matching a first object and a second object with a user input, the first object comprising one of the input action object, the output action object, the input concept object, the intermediate concept object, the other intermediate concept object and the output concept object, the second object comprising another one of the input action object, the output action object, the input concept object, the intermediate concept object, the other intermediate concept object and the output concept object;

forming an intent based on the user input;

creating a plan based on the intent and the annotated span of natural language, the plan comprising the input action object selected from a plurality of action objects that transforms the input concept object associated with the intent into the intermediate concept object, and the output action object selected from the plurality of action objects that transforms the other intermediate concept object into the output concept object associated with a goal of the intent, the other intermediate concept object comprising one of a same object as the intermediate concept object and a different object from the intermediate concept object;

executing the plan; and

outputting a value associated with the output concept object associated with the user input.

8. The computer-implemented method of claim 7, wherein creating the plan is based on a confidence score associated with the annotated span of natural language, wherein the method further comprises receiving another span of natural language annotated with another annotating object from another third-party developer, wherein creating the plan is based on the other annotated span of natural language, and wherein the other annotating object comprises a different one of the input action object, the output action object, the input concept object, the intermediate concept object, the other intermediate concept object, and the output concept object.

9. The computer-implemented method of claim 7, wherein the method further comprises outputting output a comparison based on a confidence score associated with the span of natural language annotated by the annotating object and a confidence score associated with the span of natural language without annotation by the annotating object, wherein the comparison indicates an amount of training completion relative to the span of natural language annotated with the object from the third-party developer.

10. The computer-implemented method of claim 7, wherein the method further comprises:

creating a second plan based on the intent, wherein the second plan differs from the first plan by at least one concept object and by at least one action object; and

selecting the plan for execution based on at least one of an action object cost, an action object quality, a number of planned action objects, and a confidence score associated with the annotated span of natural language.

11. The computer-implemented method of claim 7, wherein the method further comprises outputting a prompt, based on the plan, which requests an additional user input, wherein executing the plan is based on the additional user input.

12. A computer program product, comprising a non-transitory computer-readable medium having a computer-readable program code embodied therein to be executed by one or more processors, the program code including instructions to:

receive a span of natural language annotated with an annotating object from a third-party developer, the annotating object comprising one of an input action object, an output action object, an input concept object, an intermediate concept object, another intermediate concept object and an output concept object;

match a first object and a second object with a user input, the first object comprising one of the input action object, the output action object, the input concept object, the intermediate concept object, the other intermediate concept object and the output concept object, the second object comprising another one of the input action object, the output action object, the input concept object, the intermediate concept object, the other intermediate concept object and the output concept object;

form an intent based on the user input;

create a plan based on the intent and the annotated span of natural language, the plan comprising the input action object selected from a plurality of action objects that transforms the input concept object associated with the intent into the intermediate concept object, and the output action object selected from the plurality of action objects that transforms another intermediate concept object into the output concept object associated with a goal of the intent, the other intermediate concept object comprising one of a same object as the intermediate concept object and a different object from the intermediate concept object;

executing the plan; and

output a value associated with the output concept object associated with the user input.

13. The computer program product of claim 12, wherein creating the plan is based on a confidence score associated with the annotated span of natural language.

14. The computer program product of claim 12, wherein the program code further includes instructions to receive another span of natural language annotated with another annotating object from another third-party developer, wherein creating the plan is based on the other annotated span of natural language, and wherein the other annotating object comprises a different one of the input action object, the output action object, the input concept object, the intermediate concept object, the other intermediate concept object, and the output concept object.

15. The computer program product of claim 12, wherein the program code further includes instructions to output a comparison based on a confidence score associated with the span of natural language annotated by the annotating object and a confidence score associated with the span of natural language without annotation by the annotating object, wherein the comparison indicates an amount of training completion relative to the span of natural language annotated with the object from the third-party developer.

16. The computer program product of claim 12, wherein the program code further includes instructions to:

create a second plan based on the intent, wherein the second plan differs from the first plan by at least one concept object and by at least one action object; and

select the plan for execution based on at least one of an action object cost, an action object quality, a number of planned action objects, and a confidence score associated with the annotated span of natural language.

17. The computer program product of claim 12, wherein the program code further includes instructions to output a prompt, based on the plan, which requests an additional user input, wherein executing the plan is based on the additional user input.