Skill extraction system

• US 10,354,017 B2
• Filed: 11/06/2013
• Issued: 07/16/2019
• Est. Priority Date: 01/27/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Skomoroch, Peter N., Hayes, Matthew T., Gupta, Abhishek, Patil, Dhanurjay A. S.
• Primary Examiner(s): Bullock, Joshua
• Application Number:
  US14/072,955
• Publication Number:
  US 20140081928A1
• Time in Patent Office: 2,078 Days
• Field of Search: 707710
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/335 Filtering based on additional data, e.g. user or group profiles filtering in web context G06F16/9535, G06F16/9536
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06F 16/355 Class or cluster creation or modification
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 10/06398 Performance of employee with respect to a job function
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0269 based on user profile or attribute
• CPC Class Codes: G06Q 50/01 Social networking

Abstract

In an example, disclosed is a machine automated method of identifying a set of skills. In some examples, the method includes extracting a plurality of skill seed phrases from a plurality of member profiles of a social networking site, creating a plurality of disambiguated skill seed phrases by disambiguating the plurality of skill seed phrases using one or more computer processors, and de-duplicating the plurality of disambiguated skill seed phrases to create a plurality of de-duplicated skill seed phrases.

Claims

1. A computer-implemented method comprising:

causing one or more computer processors to execute instructions, the instructions causing the one or more computer processors to perform operations of;

extracting a plurality of skill seed phrases from a skills section of a plurality of member profiles of a social networking service by at least;

tokenizing data in the skills section of each of the plurality of member profiles into a plurality of tokens; and

selecting, as the plurality of skill seed phrases, tokens from the plurality of tokens that have a frequency of occurrence that is above a predetermined threshold frequency;

disambiguating the plurality of skill seed phrases to create a plurality of disambiguated skill seed phrases by at least clustering the plurality of skill seed phrases based on a count of a number of times both skills of respective pairs of the plurality of skill seed phrases are present in a same member profile of the plurality of member profiles; and

de-duplicating the plurality of disambiguated skill seed phrases to create a plurality of de-duplicated skill seed phrases, the de-duplicated skill seed phrases identifying a plurality of skills.

2. The method of claim 1, wherein extracting the plurality of skill seed phrases further comprises:

removing skill seed phrases that do not correspond with actual skills.

3. The method of claim 1 wherein the one or more computer processors comprise a computing cluster and the method is performed using a parallel and distributed process.

4. The method of claim 3, wherein the clustering is done using an expectation maximum method.

5. The method of claim 1, further comprising:

calculating attributes about the plurality of disambiguated skill seed phrases and storing the attributes in association with the plurality of disambiguated skill seed phrases.

6. The method of claim 5, wherein gathering attributes about the plurality of disambiguated skill phrases comprises computing a top industry of each of the plurality of disambiguated skill phrases based upon an industry listed in member profiles of the plurality of member profiles in which each skill seed phrase was found.

7. The method of claim 1, wherein de-duplicating the plurality of disambiguated skill seed phrases comprises:

calculating a plurality of internet search queries, at least one for each of the plurality of disambiguated skill seed phrases, the internet search queries, when executed, returning a list of internet web pages;

submitting to a crowd sourcing application, over a network, a job for each of the plurality of internet search queries for which the crowd sourcing application returns a specific internet web page selected from the list of internet web pages that is returned as a result of executing each of the plurality of the internet search queries, the returned specific internet web page representing an interpretation of a meaning of the disambiguated skill seed phrase corresponding to each internet search query; and

determining which of the plurality of disambiguated skill seed phrases identify identical skills using the returned specific internet web pages.

8. The method of claim 7, further comprising:

submitting more than one job for each of the plurality of internet search queries and using an algorithm to select among the returned specific internet web pages to determine which of the returned specific Internet web pages for each internet search query is used to determine which of the plurality of disambiguated skill seed phrases identify the identical skills.

9. The method of claim 1, wherein disambiguating the plurality of skill seed phrases to create a plurality of disambiguated skill seed phrases by at least clustering the plurality of skill seed phrases based on a count of a number of times both skills of respective pairs of the plurality of skill seed phrases are present in a same member profile of the plurality of member profiles comprises:

for each particular one of the plurality of skill seed phrases;

counting the number of times the particular one of the plurality of skill seed phrases occurred in a same member profile of the plurality of member profiles as each of the other plurality of skill seed phrases; and

calculating a probability that the particular one of the plurality of skill seed phrases is likely to appear in a same member profile as each of the plurality of other skill seed phrases; and

clustering the plurality of skill seed phrases based upon the probabilities.

10. A system comprising:

a computer processor;

a non-transitory machine readable medium comprising instructions, which when executed by the computer processor, causes the computer processor to perform operations of;

extracting a plurality of skill seed phrases from a skills section of a plurality of member profiles of a social networking service by at least;

tokenizing data in the skills section of each of the plurality of member profiles into a plurality of tokens; and

selecting tokens from the plurality of tokens with a frequency of occurrence that is above a predetermined threshold frequency as the plurality of skill seed phrases;

disambiguating the plurality of skill seed phrases to create a plurality of disambiguated skill seed phrases by at least clustering the plurality of skill seed phrases based on a count of a number of times both skills of respective pairs of the plurality of skill seed phrases are present in a same member profile of the plurality of member profiles; and

de-duplicating the plurality of disambiguated skill seed phrases to create a plurality of de-duplicated skill seed phrases, the de-duplicated skill seed phrases identifying a plurality of skills.

11. The system of claim 10, wherein the operations of extracting the plurality of skill seed phrases comprises removing skill seed phrases that do not correspond with actual skills.

12. The system of claim 10, wherein the operations comprise calculating attributes about the plurality of disambiguated skill seed phrases and storing the attributes in association with the plurality of disambiguated skill seed phrases.

13. The system of claim 12, wherein the operations of calculating attributes about the plurality of disambiguated skill phrases comprise computing a top industry of each of the plurality of disambiguated skill phrases based upon an industry listed in member profiles of the plurality of member profiles in which each skill seed phrase was found.

14. The system of claim 10, wherein the operations of de-duplicating the plurality of disambiguated skill seed phrases comprises:

calculating a plurality of internet search queries, at least one for each of the plurality of disambiguated skill seed phrases, the internet search queries, when executed, returning a list of internet web pages;

submitting to a crowd sourcing application, over a network, a job for each of the plurality of internet search queries for which the crowd sourcing application returns a specific internet web page selected from the list of internet web pages that is returned as a result of executing each of the plurality of the internet search queries, the returned specific internet web page representing an interpretation of a meaning of the disambiguated skill seed phrase corresponding to each internet search query; and

determining which of the plurality of disambiguated skill seed phrases identify identical skills using the returned specific internet web pages.

15. The system of claim 14, wherein the operations of de-duplicating the plurality of disambiguated skill seed phrase comprises:

submitting more than one job for each of the plurality of internet search queries and selecting among the returned specific internet web pages which of the returned specific internet web pages for each internet search query is used to determine which of the plurality of disambiguated skill seed phrases identify the identical skills.

16. The system of claim 10, wherein the operations of disambiguating the plurality of skill seed phrases to create a plurality of disambiguated skill seed phrases by at least clustering the plurality of skill seed phrases based on a count of the number of times both skills of respective pairs of the plurality of skill seed phrases are present in a same member profile of the plurality of member profiles comprises:

for each particular one of the plurality of skill seed phrases;

counting the number of times the particular one of the plurality of skill seed phrases occurred in a same member profile of the plurality of member profiles as each of the other plurality of skill seed phrases; and

calculating a probability that the particular one of the plurality of skill seed phrases is likely to appear in a same member profile as each of the plurality of other skill seed phrases; and

clustering the plurality of skill seed phrases based upon the calculated probabilities.

17. A non-transitory machine-readable medium that stores instructions which when performed by a machine, cause the machine to perform operations comprising:

extracting a plurality of skill seed phrases from a skills section of a plurality of member profiles of a social networking service by at least;

tokenizing data in the skills section of each of the plurality of member profiles into a plurality of tokens; and

selecting, as the plurality of skill seed phrases, tokens from the plurality of tokens that have a frequency of occurrence that is above a predetermined threshold frequency;

disambiguating the plurality of skill seed phrases to create a plurality of disambiguated skill seed phrases by at least clustering the plurality of skill seed phrases based on a count of a number of times both skills of respective pairs of the plurality of skill seed phrases are present in a same member profile of the plurality of member profiles; and

de-duplicating the plurality of disambiguated skill seed phrases to create a plurality of de-duplicated skill seed phrases;

the de-duplicated skill seed phrases identifying a plurality of skills.

18. The non-transitory machine-readable medium of claim 17, wherein the instructions for extracting the plurality of skill seed phrases comprise instructions, which when performed by the machine, cause the machine to perform the operations comprising:

removing skill seed phrases that do not correspond with actual skills.

19. The non-transitory machine-readable medium of claim 17, wherein the instructions for clustering comprise instructions, which when performed by the machine, cause the machine to perform the operations comprising:

using an expectation maximum method.

20. The non-transitory machine-readable medium of claim 17, wherein the instructions comprise instructions, which when performed by the machine, cause the machine to perform the operations of:

calculating attributes about the plurality of disambiguated skill seed phrases and storing the attributes in association with the plurality of disambiguated skill seed phrases.

21. The non-transitory machine-readable medium of claim 20, wherein the instructions for gathering attributes about the plurality of disambiguated skill phrases comprises instructions, which when performed by the machine, cause the machine to perform the operations of:

computing a top industry of each of the plurality of disambiguated skill phrases based upon an industry listed in member profiles of the plurality of member profiles in which each skill seed phrase was found.

22. The non-transitory machine-readable medium of claim 17, wherein the instructions for de-duplicating the plurality of disambiguated skill seed phrases comprises instructions, which when performed by the machine, cause the machine to perform the operations of:

calculating a plurality of internet search queries, at least one for each of the plurality of disambiguated skill seed phrases, the internet search queries, when executed, returning a list of internet web pages;

submitting to a crowd sourcing application, over a network, a job for each of the plurality of internet search queries for which the crowd sourcing application returns a specific internet web page selected from the list of internet web pages that is returned as a result of executing each of the plurality of the internet search queries, the returned specific internet web page representing an interpretation of a meaning of the disambiguated skill seed phrase corresponding to each internet search query; and

determining which of the plurality of disambiguated skill seed phrases identify identical skills using the returned specific internet web pages.

23. The non-transitory machine-readable medium of claim 22, wherein the instructions include instructions, which when performed by the machine, cause the machine to perform operations comprising:

submitting more than one job for each of the plurality of internet search queries and using an algorithm to select among the returned specific internet web pages to determine which of the returned specific internet web pages for each internet search query is used to determine which of the plurality of disambiguated skill seed phrases identify the identical skills.

24. The non-transitory machine readable medium of claim 17, wherein the operations of disambiguating the plurality of skill seed phrases to create a plurality of disambiguated skill seed phrases by at least clustering the plurality of skill seed phrases based on a count of the number of times both skills of respective pairs of the plurality of skill seed phrases are present in a same member profile of the plurality of member profiles comprises:

for each particular one of the plurality of skill seed phrases;

counting the number of times the particular one of the plurality of skill seed phrases occurred in a same member profile of the plurality of member profiles as each of the other plurality of skill seed phrases; and

calculating a probability that the particular one of the plurality of skill seed phrases is likely to appear in a same member profile as each of the plurality of other skill seed phrases; and

clustering the plurality of skill seed phrases based upon the calculated probabilities.