Skills ontology creation

• US 9,697,472 B2
• Filed: 06/04/2014
• Issued: 07/04/2017
• Est. Priority Date: 09/20/2013
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: LinkedIn Corporation (Microsoft Corporation)
• Inventors: Burgess, Matt, Skomoroch, Peter N.
• Primary Examiner(s): Holmes, Michael B
• Assistant Examiner(s): Bharadwaj, Kalpana
• Application Number:
  US14/296,258
• Publication Number:
  US 20150088793A1
• Time in Patent Office: 1,126 Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

Disclosed in some examples are systems, methods, and machine readable mediums which allow for the automatic creation of a skills hierarchy. The skills hierarchy comprises an organization of a standardized list of skills into a hierarchy that describes category relationships between the skills in the hierarchy. The category relationships may include no relationships, parent relationships, and child relationships. A skill may be considered a parent of another skill if the parent skill describes a broader category of skill that includes the child. Other relationships such as grandparent (e.g., a parent'"'"'s parent), great-grandparent, grandchild, great grandchild and so on may be defined inferentially as well. In some examples, the constructed hierarchy may be organized with broader skills at higher levels and narrower skills at lower levels.

Claims

1. A method for creating a hierarchy of skills, the method comprising:

training a skill classifier using training data, the training data comprising a plurality of skill pairs, each skill in the plurality of skill pairs selected from a predetermined list of skills;

determining for each particular skill pair in a target set of skill pairs, using the skill classifier and feature data, a probability that a first skill in each particular skill pair is a parent of a second skill in each particular skill pair, the first and second skills included in the predetermined list of skills, wherein the feature data includes a ratio of a number of times the first skill was explicitly included in member profiles of a social networking service to a number of times the first skill was implicitly included in member profiles of the social networking service;

collecting, in response to the determining the probability for each particular skill pair, a collection of probabilities for the target set of skill pairs; and

constructing a hierarchy of skills based upon the collection of probabilities for the target set of skill pairs.

2. The method of claim 1, wherein the skill classifier is a random forest classifier.

3. The method of claim 1, wherein the skill classifier takes as input second feature data describing relationships between two or more skills and training the skill classifier comprises training the skill classifier using the training data and the second feature data.

4. The method of claim 1, wherein the feature data includes data on a probability of co-occurrence in member profiles of a social networking service between the first and second skills.

5. The method of claim 4, wherein the feature data includes a closeness centrality metric based upon the probability of co-occurrence between the first and second skills.

6. The method of claim 1, wherein the feature data includes a pre-existing hierarchy describing a parent-child relationship between the first and second topics.

7. The method of claim 1, wherein the training data comprises feature data and for each particular pair of skills in the training data a designation of whether a first skill in the particular pair of skills is a parent skill of a second skill in the particular pair of skills.

8. The method of claim 7, wherein the designation is determined from hierarchy information in a free text section of a plurality of member profiles of the social networking service.

9. A non-transitory machine-readable medium for creating a hierarchy of topics, the machine-readable medium comprising instructions, which when performed by a machine, causes the machine to perform operations comprising:

training a skill classifier using training data, the training data comprising a plurality of skill pairs, each topic in the plurality of trained topic pairs selected from a predetermined list of skills;

determining for each particular skill pair in a target set of skill pairs, using the skill classifier and feature data, a probability that a first skill in each particular skill pair is a parent of a second skill in each particular skill pair, the first and second skills included in the predetermined list of skills, wherein the feature data includes a ratio of a number of times the first skill was explicitly included in member profiles of a social networking service to a number of times the first skill was implicitly included in member profiles of the social networking service;

collecting, in response to the determining the probability for each particular skill pair, a collection of probabilities for the target set of skill pairs; and

constructing a hierarchy of skills based upon the collection of probabilities for the target set of skill pairs.

10. The machine-readable medium of claim 9, wherein the skill classifier is a random forest classifier.

11. The machine-readable medium of claim 9, wherein the skill classifier takes as input second feature data describing relationships between two or more skills and the operations for training the topic classifier comprises training the topic classifier using the training data and the second feature data.

12. The machine-readable medium of claim 9, wherein the training data comprises feature data and for each particular pair of skills in the training data a designation of whether a first skill in the particular pair of skills is a parent skill of a second skill in the particular pair of skills.

13. The machine-readable medium of claim 12, wherein the designation is determined from hierarchy information in a free text section of a plurality of member profiles of the social networking service.

14. A system for creating a hierarchy of topics, the system comprising:

a processor;

a memory communicatively coupled to the processor and including instructions, when performed by the processor cause the system to perform operations to;

training a skill classifier using training data, the training data comprising a plurality of skill pairs, each skill in the plurality of skill pairs selected from a predetermined list of skills;

determining for each particular skill pair in a target set of skill pairs, using the skill classifier and feature data, a probability that a first skill in each particular skill pair is a parent of a second skill in each particular skill pair, the first and second skills included in the predetermined list of skills;

wherein the feature data includes a ratio of a number of times the first skill was explicitly included in member profiles of a social networking service to a number of times the first skill was implicitly included in member profiles of the social networking service;

collecting, in response to the determining the probability for each particular skill pair, a collection of probabilities for the target set of skill pairs; and

constructing a hierarchy of skills based upon the collection of probabilities for the target set of skill pairs.

15. The system of claim 14, wherein the skill classifier takes as input second feature data describing relationships between two or more skills and training the skill classifier comprises training the skill classifier using the training data and the second feature data.

16. The system of claim 14, wherein the feature data includes data on a probability of co-occurrence in member profiles of a social networking service between the first and second skills.

17. The system of claim 16, wherein the feature data includes a closeness centrality metric based upon the probability of co-occurrence between the first and second skills.

18. The system of claim 14, wherein the training data comprises feature data and for each particular pair of skills in the training data a designation of whether a first skill in the particular pair of skills is a parent skill of a second skill in the particular pair of skills.

19. The system of claim 18, wherein the designation is determined from hierarchy information in a free text section of a plurality of member profiles of the social networking service.