Scalable response prediction using personalized recommendation models

Patent metadata

Field	Value
Publication number	US-10102503-B2
Application number	US-201615145061-A
Country	US
Kind code	B2
Filing date	May 3, 2016
Priority date	May 3, 2016
Publication date	Oct 16, 2018
Grant date	Oct 16, 2018

Abstract

The disclosed embodiments provide a system for processing data. During operation, the system obtains a global version of a statistical model and a user-specific version of the statistical model for a user. Next, the system applies the global version to member features of the user and job features of a set of jobs to generate a first ranking of the jobs for the user. The system then applies the user-specific version to the member features and the job features for a highest-ranked subset of jobs in the first ranking to generate a second ranking of the jobs for the user. Finally, the system outputs at least a portion of the second ranking as a set of job recommendations.

First claim

What is claimed is:

1. A method, comprising:
obtaining a global version of a statistical model and a user-specific version of the statistical model for a user;
applying, by one or more computer systems, the global version to member features of the user and job features of a set of jobs to generate a first ranking of the jobs for the user;
applying, by the one or more computer systems, the user-specific version to the member features and the job features for a highest-ranked subset of jobs in the first ranking to generate a second ranking of the jobs for the user; and
outputting, to the user, at least a portion of the second ranking as a set of job recommendations;
wherein applying the global version to the member features and the job features to generate the first ranking of the jobs for the user comprises:
using a job feature key in an inverted index to retrieve the job features for a subset of the jobs that matches the job feature key; and
applying the global version to the member features and the retrieved job features to calculate a global score for each job in the subset of jobs.

2. The method of claim 1 , further comprising:
obtaining a set of job-specific models for the highest-ranked subset of jobs; and
applying the job-specific models to the member features and job features to generate the second ranking of the jobs for the user.

3. The method of claim 2 , wherein the second ranking of the jobs for the user is generated by:
for each job in the highest-ranked subset of jobs, combining a first output of the user-specific version and a second output of a job-specific model for the job to obtain a user-specific score for the job; and
re-ranking the highest-ranked subset of jobs by the user-specific score.

4. The method of claim 3 , wherein the second ranking of the jobs for the user is further generated by:
combining the first and second outputs with a third output of the global version to obtain the user-specific score for the job.

5. The method of claim 4 , wherein the third output of the global version is produced using a cross product of one or more of the member features and one or more of the job features to produce the third output.

6. The method of claim 4 , wherein the third output of the global version is produced using a similarity of one or more of the member features to one or more of the job features.

7. The method of claim 1 , wherein calculating a global score for each job in the subset of jobs comprises:
calculating the global score for the job from a weighted combination comprising a set of coefficients in the global version, the member features, and the job features of the job.

8. The method of claim 1 , further comprising:
updating the global version, the user-specific version, and one or more of the job-specific models based on one or more responses to the outputted job recommendations by the user.

9. The method of claim 1 , wherein the first and second rankings comprise an estimated probability of applying to a job by the user.

10. An apparatus, comprising:
one or more processors; and
memory storing instructions that, when executed by the one or more processors, cause the apparatus to:
obtain a global version of a statistical model and a user-specific version of the statistical model for a user;
apply the global version to member features of the user and job features of a set of jobs to generate a first ranking of the jobs for the user;
apply the user-specific version to the member features and the job features for a highest-ranked subset of jobs in the first ranking to generate a second ranking of the jobs for the user; and
output at least a portion of the second ranking as a set of job recommendations;
wherein applying the global version to the member features and the job features to generate the first ranking of the jobs for the user comprises:
using a job feature key in an inverted index to retrieve the job features for a subset of the jobs that matches the job feature key; and
applying the global version to the member features and the retrieved job features to calculate a global score for each job in the subset of jobs.

11. The apparatus of claim 10 , wherein the memory further stores instructions that, when executed by the one or more processors, cause the apparatus to:
obtain a set of job-specific models for the highest-ranked subset of jobs; and
apply the job-specific models to the member features and the job features to generate the second ranking of the jobs for the user.

12. The apparatus of claim 11 , wherein the second ranking of the jobs for the user is generated by:
for each job in the highest-ranked subset of jobs, combining a first output of the user-specific version and a second output of a job-specific model for the job to obtain a user-specific score for the job; and
re-ranking the highest-ranked subset of jobs by the user-specific score.

13. The apparatus of claim 12 , wherein the second ranking of the jobs for the user is further generated by:
combining the first and second outputs with a third output of the global version to obtain the user-specific score for the job.

14. The apparatus of claim 13 , wherein the third output of the global version is produced using a cross product of one or more of the member features and one or more of the job features to produce the third output.

15. The apparatus of claim 13 , wherein the third output of the global version is produced using a similarity of one or more of the member features to one or more of the job features.

16. The apparatus of claim 10 , wherein calculating a global score for each job in the subset of jobs comprises:
calculating the global score for the job from a weighted combination comprising a set of coefficients in the global version, the member features, and the job features of the job.

17. A system, comprising:
an analysis module comprising a non-transitory computer-readable medium comprising instructions that, when executed by a computer, cause the system to:
obtain a global version of a statistical model and a user-specific version of the statistical model for a user;
apply the global version to member features of the user and job features of a set of jobs to generate a first ranking of the jobs for the user; and
apply the user-specific version to the member features and the job features for a highest-ranked subset of jobs in the first ranking to generate a second ranking of the jobs for the user; and

a management module comprising a non-transitory computer-readable medium comprising instructions that, when executed by the computer, cause the system to output at least a portion of the second ranking as a set of job recommendations;
wherein applying the global version to the member features and the job features to generate the first ranking of the jobs for the user comprises:
using a job feature key in an inverted index to retrieve the job features for a subset of the jobs that matches the job feature key; and
applying the global version to the member features and the retrieved job features to calculate a global score for each job in the subset of jobs.

18. The system of claim 17 , wherein the non-transitory computer-readable medium of the analysis module further comprises instructions that, when executed, cause the system to:
obtain a set of job-specific models for the highest-ranked subset of jobs; and
apply the job-specific models to the member features and the job features to generate the second ranking of the jobs for the user.

19. The system of claim 17 ,

Assignees

• Linkedin Corp
• Microsoft Licensing Tech Llc

Inventors

• Zhang Xianxing
• Zhou Yitong
• Ma Yiming
• Chen Bee-Chung
• Zhang Liang
• Agarwal Deepak

Classifications

• G06Q10/1053 Primary

  Employment or hiring · CPC title

• G06F16/24578

  using ranking · CPC title

• G06F17/3053

  Physics · mapped topic