This document discusses LinkedIn's job recommendation system. It begins with an overview of LinkedIn's size and role as a professional network. It then explains that job recommendations are made implicitly based on a member's profile rather than explicit searches. The recommendation algorithm analyzes a variety of member profile features to find similar jobs. Challenges include accurately resolving entities like companies and job titles that may be referred to differently, determining member location stickiness, and leveraging the professional network. The goal is to provide relevant job recommendations to members.
3. 270+ M
Company Pages
>3M
*
Professional searches in 2012
~5.7B
90%Fortune 100 Companies
use LinkedIn to hire
*
*as of March 31, 2014
New Members joining
~2/sec
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
3
World’s largest professional network
Over 65% of members are now international
15. Corpus StatsCandidate Jobs
User Base
title
geo
company
industry
description
functional area
…
Candidate
General
expertise
specialties
education
headline
geo
experience
Current Position
title
summary
tenure length
industry
functional area
…
Similarity
(candidate expertise, job description)
0.56
Similarity
(candidate specialties, job description)
0.2
Transition probability
(candidate industry, job industry)
0.43
Title Similarity
0.8
Similarity (headline, title)
0.7
.
.
.
derived
Matching
Binary
Exact matches:
geo, industry,
…
Soft
transition
probabilities,
similarity,
…
Text
Recommendation Algorithm
Transition probabilities
Connectivity
yrs of experience to reach title
education needed for this title
…
15
Job Collection
18. ‘IBM’ has 13000+ variations
- ibm – ireland
- ibm research
- T J Watson Labs
- International Bus. Machines
Are All Companies The Same?
18
19. - Software Engineer
- Technical Yahoo
- Member Technical Staff
- Software Development Engineer
- SDE
Are All Titles The Same?
19
20. Same company with
different name
Same name but
different companies
Name Variations for IBM?
“Orion” refers to 20 diff. companies
large scale: 100M+ members, 2M+ company entities
IBM: Intl Brotherhood of Magicians
~ 13000
Challenges – Entity Resolution
20
21. § Binary classifier (LR), not
ranker
§ P({position, company
entity} is a match)
§ Features
§ Content
§ Social
§ Behavior
§ Company candidate set
leveraged from Social
graph and cosine
similarity 97% Precision
at 50% Coverage
Asonam’11, KDD’11
Challenges – Entity Resolution
21
Precision Coverage
23. § Zip code mapped to Regions
§ How sticky are those locations?
Feature Engineering – Sticky locations
23
24. § Open to relocation ?
§ Region similarity based on profiles or network
§ Region transition probability
§ Predict individuals propensity to migrate and most
likely migration target
Feature Engineering – Sticky locations
24
26. Hybrid Recommendation
Title : Research Engineer
Company : Yahoo!
Location : CA,USA
Skills : Stats, ML, Java
Title : Data Scientist
Company : Samsung
Location : PA,USA
Skills : Stats, R
Title : Analyst
Company : Microsoft
Location : CA, USA
Skills : R, ML
Title : Research Engineer <1>, Data Scientist <1>, Analyst
<1>
Company : Yahoo<1>, Samsung<1>, Microsoft<1>
Location : CA,USA <2>, PA,USA<1>
Skills : Stats<2>, ML<2>, R<2>, Java<1>
Applicant Features
Distribution
Data Scientist / Senior Data Scientist
San Jose
26
27. Information Gain
Pick Top K overrepresented features from the
applicants distribution
A representative projection of the job in the
member feature space
27
Hybrid Recommendation
28. § Why Jobs Recommendations are Different
§ Recommendation Algorithm
§ Challenges
– Entity Resolution
– Location Resolution
28
Summary