Automatic recognition of entities related to cloud incidents

Patent metadata

Field	Value
Publication number	US-2022012633-A1
Application number	US-202016925453-A
Country	US
Kind code	A1
Filing date	Jul 10, 2020
Priority date	Jul 10, 2020
Publication date	Jan 13, 2022
Grant date	—

Abstract

Systems and methods for automatic recognition of entities related to cloud incidents are described. A method, implemented by at least one processor, for processing cloud incidents related information, including entity names and entity values associated with incidents having a potential to adversely impact products or services offered by a cloud service provider is provided. The method may include using at least one processor, processing the cloud incidents related information to convert at least words and symbols corresponding to a cloud incident into machine learning formatted data. The method may further include using a machine learning pipeline, processing at least a subset of the machine learning formatted data to recognize entity names and entity values associated with the cloud incident.

First claim

What is claimed:

1 . A method, implemented by at least one processor, for processing cloud incidents related information, including entity names and entity values, the method comprising:
using the at least one processor, processing the cloud incidents related information to convert at least words and symbols corresponding to a cloud incident into machine learning formatted data; and using a machine learning pipeline, processing at least a subset of the machine learning formatted data to recognize entity names and entity values associated with the cloud incident.

2 . The method of claim 1 , further comprising using the machine learning pipeline, jointly processing at least a second subset of the machine learning formatted data with the at least the subset of the machine learning formatted data to recognize data types associated with the cloud incident.

3 . The method of claim 1 , further comprising using a multi-task learning layer, processing both the subset of the machine learning formatted data and the second subset of the machine learning formatted data to generate output data.

4 . The method of claim 3 , further comprising: (1) using a first time distributed dense layer, reshaping a first subset of the output data, wherein the first subset of the output data corresponds to entity names and entity values, to generate a first set of reshaped data and (2) using a second time distributed dense layer reshaping a second subset of the output data, wherein the second subset of the output data corresponds to data types, to generate a second set of reshaped data.

5 . The method of claim 4 , further comprising: (1) using a first attention layer, processing the first set of reshaped data, emphasizing a first set of tokens more likely to be entity names or entity types and (2) using a second attention layer, processing the second set of reshaped data, emphasizing a second set of tokens more likely to be data types.

6 . The method of claim 5 , further comprising: (1) using learned constraints associated with entity names and entity values, helping recognize the entity names and the entity values associated with the cloud incident, and (2) using learned constraints associated with data types, helping recognize the data types associated with the cloud incident.

7 . The method of claim 1 , further comprising generating a seed database of tagged entity names and tagged entity values by unsupervised tagging of entity names and entity values based on patterns extracted from cloud incidents related information.

8 . The method of claim 7 , further comprising using unsupervised label propagation of the tagged entity names and the tagged entity values, to generate training data for training the machine learning pipeline.

9 . A system, including at least one processor, for processing cloud incidents related information, including entity names and entity values associated with incidents having a potential to adversely impact products or services offered by a cloud service provider, the system configured to:
using the at least one processor, process the cloud incidents related information to convert at least words and symbols corresponding to a cloud incident into machine learning formatted data; and using a machine learning pipeline, process at least a subset of the machine learning formatted data to recognize entity names and entity values associated with the cloud incident.

10 . The system of claim 9 , further configured to jointly process at least a second subset of the machine learning formatted data with the at least the subset of the machine learning formatted data to recognize data types associated with the cloud incident.

11 . The system of claim 10 , further configured to using a multi-task learning layer, process both the subset of the machine learning formatted data and the second subset of the machine learning formatted data to generate output data.

12 . The system of claim 11 , further configured to: (1) using a first time distributed dense layer, reshape a first subset of the output data, wherein the first subset of the output data corresponds to entity names and entity values, to generate a first set of reshaped data and (2) using a second time distributed dense layer reshape a second subset of the output data, wherein the second subset of the output data corresponds to data types, to generate a second set of reshaped data.

13 . The system of claim 12 , further configured to: (1) using a first attention layer, process the first set of reshaped data, emphasizing a first set of tokens more likely to be entity names or entity types and (2) using a second attention layer, process the second set of reshaped data, emphasizing a second set of tokens more likely to be data types.

14 . The system of claim 13 , further configured to: (1) using learned constraints associated with entity names and entity values, help recognize the entity names and the entity values associated with the cloud incident, and (2) using learned constraints associated with data types, help recognize the data types associated with the cloud incident.

15 . A method, implemented by at least one processor, for processing cloud incidents related information, including entity names, entity values, and data types, the method comprising:
using the at least one processor, processing the cloud incidents related information to convert at least words and symbols corresponding to a cloud incident into machine learning formatted data; using a first machine learning pipeline, as part of a first prediction task, processing at least a first subset of the machine learning formatted data to recognize entity names and entity values associated with the cloud incident; and using a second machine learning pipeline, as part of a second prediction task, processing at least a second subset of the machine learning formatted data to recognize data types associated with the cloud incident.

16 . The method of claim 15 , further comprising using a multi-task learning layer, processing both the first subset of the machine learning formatted data and the second subset of the machine learning formatted data to generate output data.

17 . The method of claim 16 , further comprising: (1) using a first time distributed dense layer, reshaping a first subset of the output data, wherein the first subset of the output data corresponds to entity names and entity values, to generate a first set of reshaped data and (2) using a second time distributed dense layer reshaping a second subset of the output data, wherein the second subset of the output data corresponds to data types, to generate a second set of reshaped data.

18 . The method of claim 17 , further comprising: (1) using a first attention layer, processing the first set of reshaped data, emphasizing a first set of tokens more likely to be entity names or entity types and (2) using a second attention layer, processing the second set of reshaped data, emphasizing a second set of tokens more likely to be data types.

19 . The method of claim 18 , further comprising: (1) using learned constraints associated with entity names and entity values, helping recognize the entity names and the entity values associated with the cloud incident, and (2) using learned constraints associated with data types, helping recognize the data types associated with the cloud incident.

20 . The method of claim 15 , further comprising: (1) generating a seed database of tagged entity names and tagged entity values by unsupervised tagging of entit

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Molahalli Manish Shetty
• Bansal Chetan
• Kumar Sumit
• Rao Nikitha
• Nagappan Nachiappan
• Zimmermann Thomas Michael Josef

Classifications

• G06F18/2155

  characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling · CPC title

• G06Q10/103 Primary

  Workflow collaboration or project management · CPC title

• H04L41/50

  Network service management, e.g. ensuring proper service fulfilment according to agreements · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06F40/295

  Named entity recognition · CPC title