Method and system to determine a category score of a social network member

• US 9,418,119 B2
• Filed: 02/18/2015
• Issued: 08/16/2016
• Est. Priority Date: 03/25/2014
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: LinkedIn Corporation (Microsoft Corporation)
• Inventors: Liu, Yan, Liu, Shaobo
• Primary Examiner(s): Ly, Anh
• Application Number:
  US14/625,424
• Publication Number:
  US 20150278218A1
• Time in Patent Office: 545 Days
• Field of Search: 707/722, 707/723, 707/731, 707/732, 707/750, 707/755, 707/692, 707/738, 707/748, 707/758, 707/779, 707/798, 707/710, 707/E17.108, 707/E17.014, 707/E17.084, 707/E17.054, 705/7.29, 705/7.14, 705/7.34, 705/14.45, 705/26.1, 705/321, 705/14.66, 705/319, 709/203, 709/204, 709/206, 709/217, 709/218, 709/219, 709/223, 709/224, 709/226, 709/230, 709/231, 715/808
• US Class Current: 1/1
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06Q 50/01 Social networking

Abstract

A method and system to determine a category score of a social network member is described. An example system comprises a sample selector, a weight value module, a storing module, an access module, and a category score module. The sample selector selects a sample of member profiles from the profiles maintained by an on-line social network system. The weight value module obtains respective weight values associated with various phrases present in the sample of member profiles. The access module accesses a member profile and the weighted phrases associated with a certain category. The category score module determines a category score for the member profile based on a presence of one or more phrases from the plurality of weighted phrases in the member profile.

Claims

1. A method comprising:

using at least one processor coupled to a memory, selecting a seed sample of member profiles in an on-line social network system, a member profile from the member profiles representing a member of the on-line social network system, the on-line social network system maintaining one or more member categories, the seed sample of member profiles comprising a plurality of phrases, each member profile from the member profiles comprising two or more phrases;

associating each profile from the seed sample of member profiles with a target category from the one or more member categories;

for each phrase from the plurality phrases, obtaining a weight value utilizing regularized linear regression, a weight value of a phrase from the plurality phrases calculated based on presence or absence of a respective phrase in the seed sample of member profiles, a combination of a phrase from the plurality phrases and its weight value comprising a weighted phrase, the plurality of phrases with their respective weight values comprising a plurality of weighted phrases for the target category, the obtaining of a weight value for each phrase from the plurality phrases comprises identifying a neutral phrase from the plurality phrases and assigning a zero weight value to the neutral phrase;

accessing a member profile from the member profiles and the plurality of weighted phrases;

based on a presence of one or more phrases from the plurality of weighted phrases in the member profile, generating a category score for the member profile, the category score indicating a likelihood of the member profile being associated with the target category;

retrieving, from a database, the category score of the member profile;

comparing the category score of the member profile to a threshold value; and

based on a result of the comparing, selectively identifying the member profile as associated with the target category.

2. The method of claim 1, wherein each profile from the seed sample of member profiles comprises data indicative of a purchased subscription to a service provided by the on-line social networking system.

3. The method of claim 1, comprising assigning a negative weight value to a phrase from the plurality phrases based on the absence of the phrase in the seed sample of member profiles.

4. The method of claim 1, comprising assigning a weight value to a phrase from the plurality phrases based on frequency of the phrase in the seed sample of member profiles.

5. The method of claim 1, wherein the obtaining of a weight value for each phrase from the plurality phrases comprises utilizing support vector machine (SVM).

6. The method of claim 1, wherein the obtaining of a weight value for each phrase from the plurality phrases comprises utilizing Naï

ve Bayes algorithm.

7. The method of claim 1, comprising:

responsive to identifying the member profile as associated with the target category, sending a communication to a member represented by the member profile.

8. A computer-implemented system comprising:

a sample selector, implemented using at least one processor, to select, using the at least one processor, a random sample of member profiles in an on-line social network system, a member profile from the member profiles representing a member of the on-line social network system, the on-line social network system maintaining one or more member categories, the random sample of member profiles comprising a plurality of phrases, each member profile from the member profiles comprising two or more phrases;

a weight value module, implemented using at least one processor, to obtain, using regularized linear regression, for each phrase from the plurality phrases, a weight value of a phrase from the plurality phrases calculated based on presence or absence of a respective phrase in the seed sample of member profiles, a combination of a phrase from the plurality phrases and its weight value comprising a weighted phrase, the plurality of phrases with their respective weight values comprising a plurality of weighted phrases for the target category, the weight value identify a neutral phrase from the plurality phrases and assign a zero weight value to the neutral phrase;

an access module, implemented using at least one processor, to access a member profile from the member profiles and the plurality of weighted phrases, using the at least one processor;

a category score module, implemented using at least one processor, to determine, using the at least one processor, based on a presence of one or more phrases from the plurality of weighted phrases in the member profile, a category score for the member profile, the category score indicating a likelihood of the member profile being associated with the target category; and

a member segmentation module to;

retrieve, from a database, the category score of the member profile;

compare the category score of the member profile to a threshold value; and

based on a result of the comparing, selectively identify the member profile as associated with the target category.

9. The system of claim 8, wherein each profile from the seed sample of member profiles comprises data indicative of a purchased subscription to a service provided by the on-line social networking system.

10. The system of claim 8, wherein the weight value module is to assign a negative weight value to a phrase from the plurality phrases based on the absence of the phrase in the seed sample of member profiles.

11. The system of claim 8, wherein the weight value module is to assign a weight value to a phrase from the plurality phrases based on frequency of the phrase in the seed sample of member profiles.

12. The system of claim 8, wherein the weight value module is to utilize support vector machine (SVM).

13. The system of claim 8, wherein the weight value module is to utilize Naï

ve Bayes algorithm.

14. A machine-readable non-transitory storage medium having instruction data to cause a machine to perform operations comprising:

selecting a seed sample of member profiles in an on-line social network system, a member profile from the member profiles representing a member of the on-line social network system, the on-line social network system maintaining one or more member categories, the seed sample of member profiles comprising a plurality of phrases, each member profile from the member profiles comprising two or more phrases;

for each phrase from the plurality phrases, obtaining a weight value utilizing regularized linear regression, a weight value of a phrase from the plurality phrases calculated based on presence or absence of a respective phrase in the seed sample of member profiles, a combination of a phrase from the plurality phrases and its weight value comprising a weighted phrase, the plurality of phrases with their respective weight values comprising a plurality of weighted phrases for the target category, the obtaining of a weight value for each phrase from the plurality phrases comprises identifying a neutral phrase from the plurality phrases and assigning a zero weight value to the neutral phrase;

accessing a member profile from the member profiles and the plurality of weighted phrases; and

determining, based on a presence of one or more phrases from the plurality of weighted phrases in the member profile, a category score for the member profile, the category score indicating a likelihood of the member profile being associated with the target category;

retrieving, from a database, the category score of the member profile;

comparing the category score of the member profile to a threshold value; and

based on a result of the comparing, selectively identifying the member profile as associated with the target category.