Determining rationale of cognitive system output

Patent metadata

Field	Value
Publication number	US-11037049-B2
Application number	US-201816173603-A
Country	US
Kind code	B2
Filing date	Oct 29, 2018
Priority date	Oct 29, 2018
Publication date	Jun 15, 2021
Grant date	Jun 15, 2021

Abstract

According to one or more embodiments of the present invention, a computer-implemented method includes generating, by a cognitive system, an answer for a user-provided query using an analytics algorithm. The answer is based on a set of data sources. The method further includes determining an influence weightage of each data source from the set of data sources. The method further includes generating and presenting a rationale for the answer based on the influence weightage.

First claim

What is claimed is:

1. A computer-implemented method comprising:
generating, by a cognitive system, an answer for a user-provided query using an analytics algorithm, the answer is based on a set of data sources;
determining an influence weightage of each data source from the set of data sources;
extracting text from at least one data source of the set of data sources;
generating a rationale supporting the answer from at least one data source of the set of data sources, wherein the rationale comprises the text;
combining the answer with the rationale based on the influence weightage; and
generating and presenting the combined answer and rationale in a single cell of a user interface,
wherein determining the influence weightage of each data source on the answer comprises back propagating to adjust the influence weightage of each operation performed in the analytics algorithm to obtain the answer, and
wherein the rationale comprises a visual representation of the set of data sources and one or more operations performed to generate the answer.

2. The computer-implemented method of claim 1 , wherein the rationale further comprises a summary of each data source from the set of data sources used to generate the answer.

3. The computer-implemented method of claim 1 , wherein a data source from the set of data sources is one from a group of electronic data sources comprising books, articles, journal papers, social media posts, and blog posts.

4. The computer-implemented method of claim 1 , wherein the answer is a first answer, the method further comprising:
adjusting, by a user, the influence weightage of a first data source from the set of data sources; and
generating, by the cognitive system, a second answer for the user-provided query using the analytics algorithm, the answer is based on the set of data sources using the adjusted influence weightage.

5. The computer-implemented method of claim 4 , wherein adjusting the influence weightage can omit the first data source from being used to generate the answer.

6. A system comprising:
a user interface;
a memory having computer readable instructions; and
one or more processors for executing the computer readable instructions, the computer readable instructions controlling the one or more processors to perform operations comprising:
generating an answer for a user-provided query using machine learning, the answer is based on a set of data sources;
extracting text from at least one data source of the set of data sources;
determining an influence weightage of each data source from the set of data sources, and generating a rationale for the answer based on the influence weightage, wherein the rationale comprises the text;
combining the answer with the rationale based on the influence weightage; and
presenting, by the use interface, the combined answer and rationale to a user in a single cell of the user interface,
wherein determining the influence weightage of each data source on the answer comprises back propagating to adjust the influence weightage of each operation performed in the analytics algorithm to obtain the answer, and
wherein the rationale comprises a visual representation of the set of data sources and one or more operations performed to generate the answer.

7. The system of claim 6 , wherein the rationale further comprises a summary of each data source from the set of data sources used to generate the answer.

8. The system of claim 6 , wherein a data source from the set of data sources is one from a group of electronic data sources comprising books, articles, journal papers, social media posts, and blog posts.

9. The system of claim 6 , wherein the answer is a first answer, the method further comprising:
adjusting, by the user interface, the influence weightage of a first data source from the set of data sources; and
generating, by the answer generator, a second answer for the user-provided query using the machine learning, the answer is based on the set of data sources using the adjusted influence weightage.

10. The system of claim 9 , wherein adjusting the influence weightage can omit the first data source from being used to generate the answer.

11. A computer program product comprising a computer readable storage medium having stored thereon program instructions executable by one or more processing devices to perform a method comprising:
generating, by a cognitive system, an answer for a user-provided query using machine learning, the answer is based on a set of data sources;
determining an influence weightage of each data source from the set of data sources;
extracting text from at least one data source of the set of data sources;
generating a rationale for the answer based on the influence weightage, wherein the rationale comprises the text;
combining the answer with the rationale based on the influence weightage; and
presenting the combined answer and rationale to a user in a single cell of a user interface,
wherein determining the influence weightage of each data source on the answer comprises back propagating to adjust the influence weightage of each operation performed in the analytics algorithm to obtain the answer, and
wherein the rationale comprises a visual representation of the set of data sources and one or more operations performed to generate the answer.

12. The computer program product of claim 11 , wherein the rationale further comprises a summary of each data source from the set of data sources used to generate the answer.

13. The computer program product of claim 11 , wherein a data source from the set of data sources is one from a group of electronic data sources comprising books, articles, journal papers, social media posts, and blog posts.

14. The computer program product of claim 11 , wherein the answer is a first answer, the method further comprising:
adjusting, by a user, the influence weightage of a first data source from the set of data sources; and
generating, by the cognitive system, a second answer for the user-provided query using the machine learning, the answer is based on the set of data sources using the adjusted influence weightage.

Assignees

• IBM

Inventors

• Chan Yuk L
• Flom Mikhail
• Jumba Albert S
• Kumar Niraj
• Luthra Tejinder
• Mallepalle Sue
• Pistoleanu Florin-Traian
• Ravindranath Goduwin R
• Sreedharan Rekha M
• Sweiss Abraham
• Taylor Sheryl
• Yarlagadda Hemanth

Classifications

• G06N3/042

  Knowledge-based neural networks; Logical representations of neural networks · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N5/02

  Knowledge representation; Symbolic representation · CPC title

• G06N5/041 Primary

  Abduction · CPC title

• G06N5/045

  Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence · CPC title