Cognitive distributed network

• US 10,713,574 B2
• Filed: 04/10/2014
• Issued: 07/14/2020
• Est. Priority Date: 04/10/2014
• Status: Active Grant

• Current Assignee: Kyndryl Incorporated
• Original Assignee: International Business Machines Corporation
• Inventors: Baughman, Aaron K., Harrison, Thomas B., O'Connell, Brian M., Pearthree, Herbert D.
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): Smith, Brian M
• Application Number:
  US14/249,679
• Publication Number:
  US 20150294216A1
• Time in Patent Office: 2,287 Days
• Field of Search: None
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 40/279 Recognition of textual entities
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/126 Evolutionary algorithms, e.g. genetic algorithms or genetic programming
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

Approaches are provided for answering an inquiry of a cognitive distributed network. An approach includes receiving the inquiry at the cognitive distributed network. The approach further includes determining a classification for the inquiry based on natural language of the inquiry. The approach further includes classifying the inquiry as a single question class. The approach further includes determining, by at least one computing device, a type of introspection to be used by the cognitive distributed network on the inquiry. The approach further includes generating an answer to the inquiry based on the determined type of introspection.

Claims

1. A method for answering an inquiry of a cognitive distributed computer network, comprising:

generating, by a computing device, an evidence based ranked hypothesis for a training inquiry;

comparing, by the computing device, the evidence based ranked hypothesis for the training inquiry with an actual answer from a key;

minimizing errors between the evidence based ranked hypothesis for the training inquiry with the actual answer from the key by using a Newton Raphson learning algorithm and changing weights of each type of feature value within a logistic regression algorithm;

receiving the inquiry at an introspective module of the computing device in the cognitive distributed computer network;

determining, by the computing device, a classification for the inquiry based on natural language of the inquiry;

classifying, by the computing device, the inquiry as a single question class;

determining, by the introspective module of the computing device, a type of introspection to be used by the cognitive distributed computer network on the inquiry to generate an answer to the inquiry, the type of introspection being based on an amount of detail provided in the inquiry;

suggesting, by the computing device, related terms which are related to the inquiry based on the amount of detail provided in the inquiry;

adjusting, by the computing device, a threshold of a precision oriented introspection algorithm that has the determined type of introspection for generating the answer to the inquiry by using related terms, from a previous inquiry, outside a predetermined threshold of the precision oriented introspection algorithm;

minimizing, by the computing device, false positive responses using the adjusted threshold of the precision oriented introspection algorithm;

generating, by the computing device, the answer to the inquiry using natural text and a cognitive cloud visualization which comprises a graphical chart that shows a predictive cloud using unstructured information management architecture based on the determined type of introspection, minimizing errors between the evidence based ranked hypothesis for the training inquiry with the actual answer from the key, and minimizing a number of false positive responses obtained using the adjusted threshold of the precision oriented introspection algorithm; and

provisioning and allocating cloud computing resources for the cognitive distributed computer network based on the received inquiry, the generated answer to the inquiry, and the predictive cloud using unstructured information management architecture by applying predictive analytics and forecasting.

2. The method of claim 1, wherein the inquiry is received in response to a human wanting to discover a health of the cognitive distributed network through a visual interface that takes input from the human using the natural language.

3. The method of claim 1, wherein the determining the classification for the inquiry is based on key words and phrases within the natural language of the inquiry such that the inquiry is classified as the single question class in response to the inquiry including predetermined phrases or the inquiry is classified as a conversational class in response to the inquiry pertaining to a string of inquiries.

4. The method of claim 1, wherein the determined type of introspection is selected from the group consisting of recall, precision and average, and the determined type of introspection is used to analyze, formulate, and generate the answer to the inquiry using a collection of information and resources available to the computing device.

5. The method of claim 1, wherein a service provider at least one of creates, maintains, deploys and supports a computer infrastructure that includes the computing device.

6. The method of claim 1, wherein steps of claim 1 are provided by a service provider on at least one of a subscription, advertising, and fee basis.

7. The method of claim 1, further comprising a forecasting module which uses symbiosis to predict fault tolerance and disk failures, and minimizing false negative responses using a midterm introspection algorithm, projecting to a middle, and replying to additional information using a past or predicted performance of the cognitive distributed computer network in view of the projected middle.

8. A computer program product for answering an inquiry of a cognitive distributed computer network, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable and executable by a computing device to cause the computing device to:

generate, by the computing device, an evidence based ranked hypothesis for a training inquiry;

compare, by the computing device, the evidence based ranked hypothesis for the training inquiry with an actual answer from a key;

minimize errors between the evidence based ranked hypothesis for the training inquiry with the actual answer from the key by using a Newton Raphson learning algorithm and changing weights of each type of feature value within a logistic regression algorithm;

receive, by an introspective module of the computing device in the cognitive distributed computer network, the inquiry;

determine, by the computing device, a classification for the inquiry;

classify, by the computing device, the inquiry as a conversational class;

determine, by the computing device, whether the inquiry pertains to past performance or future performance of the cognitive distributed computer network;

in response to determining the inquiry pertains to the past performance, apply, by the computing device, natural language processing to the inquiry to determine how the cognitive distributed computer network was performing in the past using an ontology mapping module by mapping introspection tags to analytics correlating at least to a natural language;

in response to determining the inquiry pertains to the future performance, apply, by the computing device, the natural language processing to the inquiry to determine how the cognitive distributed computer network will be performing in the future using a forecasting module which simulates ahead of horizon metrics and forecasting by applying the natural language processing to questions or conversation correspondence;

suggest, by the computing device, related terms which are related to the inquiry based on the amount of detail provided in the inquiry;

adjust, by the computing device, a threshold of a recall oriented introspection algorithm by using related terms for generating a reply to the inquiry, from a previous inquiry, outside a predetermined threshold of the recall oriented introspection algorithm;

use, by the computing device, the adjusted threshold of the recall oriented introspection algorithm to minimize false negative responses;

reply, by the computing device, to the inquiry using natural text and a cognitive cloud visualization which comprises a graphical chart that shows how a predictive cloud is using unstructured information management architecture based on the determination of how the cognitive distributed computer network was performing in the past or will be performing in the future, minimized errors between the evidence based ranked hypothesis for the training inquiry with the actual answer from the key, and the use of the adjusted threshold of the recall oriented introspection algorithm; and

provision and allocate cloud computing resources for the cognitive distributed computer network based on the received inquiry, the reply to the inquiry, and the predictive cloud using unstructured information management architecture by applying predictive analytics and forecasting,

wherein the key comprises a training set.

9. The computer program product of claim 8, wherein the determining whether the inquiry pertains to the past performance or the future performance is based on natural language of the inquiry and a temporal account of events discussed therein.

10. The computer program product of claim 8, wherein in response to determining the inquiry pertains to the past performance, use, by the computing device, distributed network based ontology to extract concepts, determine correspondences between the concepts, map introspection tags to analytics that measure health and performance of the cognitive distributed computer network, and retrieve the concepts in response to the inquiry.

11. The computer program product of claim 10, wherein the mapped introspection tags are correlated to natural language words, focus, and entities.

12. The computer program product of claim 8, wherein in response to determining the inquiry pertains to the future performance, apply, by the computing device, forecasting based algorithms that use time series autoregressive integrated moving average methods to simulate ahead of the horizon health metrics and performance metrics.

13. The computer program product of claim 12, wherein the forecasting based algorithms are configured to simulate ahead of time horizon forecasting.

14. The computer program product of claim 8, wherein in response to determining the inquiry pertains to the future performance, use, by the computing device, symbiosis to predict fault tolerance and disk failures for health and performance forecasting of the cognitive distributed computer network.

15. The computer program product of claim 8, further comprising relax, by the computing device, a threshold of natural language processing questions to distributed network ontology relationships to minimize false positive responses, and the distributed network ontology relationships comprise mapped introspection tags to analytics that measure health and performance of the cognitive distributed computer network.

16. The computer program product of claim 8, further comprising receive, by the computing device, additional information provided in response to the reply to the inquiry and a reaction to the determined performance of the cognitive distributed computer network in the past, wherein the additional information includes additional natural language to narrow a scope of the inquiry and is received in response to a human reacting to how the cognitive distributed computer network was performing in the past.

17. The computer program product of claim 16, further comprising:

determine, by the computing device, whether the additional information pertains to the past performance or the future performance of the cognitive distributed computer network;

in response to determining the additional information pertains to the past performance, apply, by the computing device, the natural language processing to the additional information to determine how the cognitive distributed computer network was performing in the past; and

in response to determining the additional information pertains to the future performance, apply, by the computing device, the natural language processing to the additional information to determine how the cognitive distributed computer network will be performing in the future.

18. The computer program product of claim 17, further comprising use, by the computing device, a midterm introspection algorithm to minimize false positive responses and false negative responses and projecting to a middle.

19. The computer program product of claim 18, further comprising reply, by the computing device, to the additional information using another natural text in the cognitive cloud visualization that shows how the predictive cloud is using unstructured information management architecture based on the determination of how the cognitive distributed computer network was performing in the past or will be performing in the future, and the use of the midterm introspection algorithm.

20. A system for answering an inquiry of a cognitive distributed computer network, comprising:

a CPU, a computer readable memory and a computer readable storage medium;

program instructions to generate an evidence based ranked hypothesis for a training inquiry;

program instructions to compare the evidence based ranked hypothesis for the training inquiry with an actual answer from a key;

program instructions to minimize errors between the evidence based ranked hypothesis for the training inquiry with the actual answer from the key by using a Newton Raphson learning algorithm and changing weights of each type of feature value within a logistic regression algorithm;

program instructions to receive the inquiry at an introspective module of a computing device in the cognitive distributed computer network;

program instructions to determine a classification for the inquiry using the computing device;

program instructions to classify the inquiry as a conversational class using the computing device;

program instructions to determine whether the inquiry pertains to past performance or future performance of the cognitive distributed computer network;

in response to determining the inquiry pertains to the past performance, program instructions to apply natural language processing to the inquiry to determine how the cognitive distributed computer network was performing in the past using an ontology mapping module;

in response to determining the inquiry pertains to the future performance, program instructions to the apply natural language processing to the inquiry to determine how the cognitive distributed computer network will be performing in the future using a forecasting module;

program instructions to suggest, by the computing device, related terms which are related to the inquiry based on the amount of detail provided in the inquiry;

program instructions to adjust a threshold of a recall oriented introspection algorithm by using related terms for generating a reply to the inquiry, from a previous inquiry, outside a predetermined threshold of the recall oriented introspection algorithm;

program instructions to use the adjusted threshold of the recall oriented introspection algorithm to minimize false negative responses using the computing device;

program instructions to reply to the inquiry using natural text and a cognitive cloud visualization which comprises a graphical chart that shows a predictive cloud using unstructured information management architecture based on the determination of how the cognitive distributed computer network was performing in the past or will be performing in the future using the computing device, minimized errors between the evidence based ranked hypothesis for the training inquiry with the actual answer from the key, and the use of the adjusted threshold of the recall oriented introspection algorithm;

program instructions to provision and allocate cloud computing resources for the cognitive distributed computer network based on the received inquiry, the reply to the inquiry, and the predictive cloud using unstructured information management architecture by applying predictive analytics and forecasting; and

program instructions to update the unstructured information management architecture in response to new features being added to unstructured information management architecture pipelines,

wherein the program instructions are stored on the computer readable storage medium for execution by the CPU via the computer readable memory.