INFERRING PROFESSIONAL REPUTATIONS OF SOCIAL NETWORK MEMBERS

• US 20160292643A1
• Filed: 06/30/2015
• Published: 10/06/2016
• Est. Priority Date: 03/31/2015
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: LinkedIn Corporation (Microsoft Corporation)
• Inventors: Puett, Douglas Jobe, Rodriguez, Mario Sergio
• Granted Patent: US 9,792,588 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 10/1053 Employment or hiring
• CPC Class Codes: G06Q 50/01 Social networking

Abstract

Techniques for inferring a professional reputation for a member of an online social networking service are described. A recommendation graph is generated from professional recommendations submitted by members of the online social networking service for other members of the online social networking service. Using the generated recommendation graph, a reputation model is trained to learn a respective importance for each respective feature of a set of features of electronic profiles on the online social networking service. A professional reputation of a member of the online social networking service is estimated by applying the trained reputation model to a feature vector of the electronic profile of the member, producing a score representing the professional reputation of the member.

Claims

1. A method comprising:

using one or more computer processors to perform operations of;

generating, for an online social networking service having members with individual electronic profiles having features, a recommendation graph including a plurality of vertices and a plurality of edges, each vertex representing a member, each edge representing a recommendation of a recommendee member by a recommender member, the recommendation accepted by the recommendee member;

training a reputation model to learn a respective importance for each respective feature of a subset of features of the electronic profiles, by providing the generated recommendation graph to a classifier;

estimating the professional reputation of a member by applying the trained reputation model to a feature vector of the member, the feature vector including features included in the subset of features, wherein applying includes adjusting a respective feature value in the feature vector of the member by a respective weight corresponding to the respective learned importance of the respective feature; and

aggregating a set of estimated professional reputations of members that have engaged with a content item posted on the online social networking service;

determining, based on the aggregated set of estimated professional reputations, whether the posted content item is spam content; and

generating an updated user interface to include;

the posted content item;

a first visual indictor of the aggregated set of estimated professional reputations as an aggregated reputation; and

a second visual indicator of whether the posted content item is spam content.

2. The method of claim 1, wherein an edge in the plurality of edges includes a label designated by the recommender member and accepted by the recommendee member;

wherein the label describes a professional relationship between the recommender member and the recommendee member; and

wherein the label is one of “

managed recommendee directly”

, “

reported directly to recommendee”

, “

managed recommendee indirectly”

, “

reported indirectly to recommendee”

, “

senior to recommendee, but not in recommendee'"'"'s chain of command”

, “

junior to recommendee, but recommendee not in chain of command”

, “

worked with recommendee in the same group”

, and “

worked with recommendee in different groups”

.

3. The method of claim 1, wherein the subset of features for an electronic profile of a member of the online social networking service includes at least one of age, gender, industry, job title, location, employer, employer size, position in the recommendation graph, title-based seniority, title-based seniority of similar members, length of professional career, length of membership in the online social networking service, and degree obtained.

4. The method of claim 1, wherein the training includes providing a plurality of recommender-recommendee vertex pairs to the classifier, each recommender-recommendee vertex pair in the plurality connected by an edge in the recommendation graph and including features of the recommender member and features of the recommendee member.

5. The method of claim 1, wherein the training includes providing a plurality of feature vectors to the classifier, each respective feature vector including:

features of a respective recommender member and features of a respective recommendee member, the respective recommender member connected by an edge to the respective recommendee member; and

features of interactions between the respective recommender member and the respective recommendee member.

6. The method of claim 1, wherein the estimated professional reputation of the member is a level of seniority of the member.

7. The method of claim 1, wherein the plurality of vertices of the generated recommendation graph includes only members having an electronic profile on the online social networking service indicating involvement in a particular industry.

8. The method of claim 1, wherein the estimate of the professional reputation of the member is a numerical score.

9. The method of claim 1, wherein training includes projecting the generated recommendation graph onto one dimension, the projecting performed using at least one of principal component analysis and the PageRank algorithm.

11. The method of claim 9, wherein projecting includes the classifier using logistic regression to project the generated recommendation graph onto one dimension.

10. (canceled)

12. A system comprising:

a machine-readable medium including machine-readable instructions which, when executed by a processor of a machine, cause the machine to perform operations comprising;

generating, for an online social networking service having members with individual electronic profiles having features, a recommendation graph including a plurality of vertices and a plurality of edges, each vertex representing a member, each edge representing a recommendation of a recommendee member by a recommender member, the recommendation accepted by the recommendee member;

training a reputation model to learn a respective importance for each respective feature of a subset of features of the electronic profiles, by providing the generated recommendation graph to a classifier;

estimating the professional reputation of a member by applying the trained reputation model to a feature vector of the member, the feature vector including features included in the subset of features, wherein applying includes adjusting a respective feature value in the feature vector of the member by a respective weight corresponding to the respective learned importance of the respective feature; and

aggregating a set of estimated professional reputations of members that have engaged with a content item posted on the online social networking service;

determining, based on the aggregated set of estimated professional reputations, whether the posted content item is spam content; and

generating an updated user interface to include;

the posted content item;

a first visual indictor of the aggregated set of estimated professional reputations as an aggregated reputation; and

a second visual indicator of whether the posted content item is spam content.

13. The system of claim 12, wherein an edge in the plurality of edges includes a label designated by the recommender member and accepted by the recommendee member;

wherein the label describes a professional relationship between the recommender member and the recommendee member; and

wherein the label is one of “

managed recommendee directly”

, “

reported directly to recommendee”

, “

managed recommendee indirectly”

, “

reported indirectly to recommendee”

, “

senior to recommendee, but not in recommendee'"'"'s chain of command”

, “

junior to recommendee, but recommendee not in chain of command”

, “

worked with recommendee in the same group”

, and “

worked with recommendee in different groups”

.

14. The system of claim 12, wherein the subset of features for an electronic profile of a member of the online social networking service includes at least one of age, gender, industry, job title, location, employer, employer size, position in the recommendation graph, title-based seniority, title-based seniority of similar members, length of professional career, length of membership in the online social networking service, and degree obtained.

15. The system of claim 12, wherein the estimated professional reputation of the member is a level of seniority of the member.

16. The system of claim 12, wherein the plurality of vertices of the generated recommendation graph includes only members having an electronic profile on the online social networking service indicating involvement in a particular industry.

17. The system of claim 12, wherein the estimate of the professional reputation of the member is a numerical score.

18. The system of claim 12, wherein training includes projecting the generated recommendation graph onto one dimension, the projecting performed using at least one of principal component analysis and the PageRank algorithm.

19. A non-transitory machine-readable storage medium including instructions, which, when executed by a processor of a machine, cause the machine to perform operations comprising:

generating, for an online social networking service having members with individual electronic profiles having features, a recommendation graph including a plurality of vertices and a plurality of edges, each vertex representing a member, each edge representing a recommendation of a recommendee member by a recommender member, the recommendation accepted by the recommendee member;

training a reputation model to learn a respective importance for each respective feature of a subset of features of the electronic profiles, by providing the generated recommendation graph to a classifier;

estimating the professional reputation of a member by applying the trained reputation model to a feature vector of the member, the feature vector including features included in the subset of features, wherein applying includes adjusting a respective feature value in the feature vector of the member by a respective weight corresponding to the respective learned importance of the respective feature; and

aggregating a set of estimated professional reputations of members that have engaged with a content item posted on the online social networking service;

determining, based on the aggregated set of estimated professional reputations, whether the posted content item is spam content; and

generating an updated user interface to include;

the posted content item;

a first visual indictor of the aggregated set of estimated professional reputations as an aggregated reputation; and

a second visual indicator of whether the posted content item is spam content.

20. The non-transitory machine-readable storage medium of claim 19, wherein an edge in the plurality of edges includes a label designated by the recommender member and accepted by the recommendee member;

wherein the label describes a professional relationship between the recommender member and the recommendee member; and

wherein the label is one of “

managed recommendee directly”

, “

reported directly to recommendee”

, “

managed recommendee indirectly”

, “

reported indirectly to recommendee”

, “

senior to recommendee, but not in recommendee'"'"'s chain of command”

, “

junior to recommendee, but recommendee not in chain of command”

, “

worked with recommendee in the same group”

, and “

worked with recommendee in different groups”

.