Data-intensive Bidirectional Matching of
Job Seeker to Vacancy
Sisay Adugna Chala
Institute of Knowledge Based Systems and Knowledge Management
University of Siegen
sisay.chala@uni-siegen.de
June 26, 2017
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 1 / 20
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 2 / 20
Motivation
Jobs are not filled because of lack of the right applicant 1
Job seekers don’t have access to the right jobs
Vacancies are not complete
Applicants do not have complete information, e.g., preferences
Jobs are not filled because of applicants quitting, being fired 2
Applicants under- or overstated their suitability to the job
1 (Belloni, M., Brugiavini, A., Meschi, E., and Tijdens, K. G., 2016)
2 (Branham, L., 2012)
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 3 / 20
Motivation ...
Huge online data of job descriptions entered by job seekers and
job holders
Employers produce large volume of vacancy data online
Due to this volume, not all vacancies are reachable by job-seekers
who have relevant skill set
Not all vacancies specify all the required skills
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 4 / 20
Motivation ...
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 5 / 20
Matching
Why is matching of job seeker to vacancy is challenging as
compared to other matching problems?
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 6 / 20
Bidirectional Matching
Definition
Bidirectional Matching measures
the degree of semantic similarity of
job seekers against vacancies and
matches one to the other
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 7 / 20
Methods and Tools
Web mining – to scrap online vacancy and resumé data
Natural language processing – to represent the textual data
Machine learning – to extract, and present the analysis result
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 8 / 20
Data Source
Wageindicator dataset
Online job description data
from occupational standards
Online vacancy data via
crawling
Online resumé
Social networking data
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 9 / 20
Data Pre-processing
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 10 / 20
Document Similarity Analysis and Clustering
Building a document vector in order to represent the document as
a whole using statistically most important words contained in
the document
Similarity analysis applications work by comparing the vectors of
documents using Latent Semantic Analysis (LSA)
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 11 / 20
Baseline
c000 Software Developer Intern
c001 Business Development Manager IOT
c002 Web Developer
v000 Software Development and Integration Manager
v001 Software Development Manager, IT integration
v002 Industrial Internet of Things Manager
v003 IoT Software Engineer, v004 IoT Saas Architect
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 12 / 20
Baseline
c000 Software Developer Intern
c001 Business Development Manager IOT
c002 Web Developer
v000 Software Development and Integration Manager
v001 Software Development Manager, IT integration
v002 Industrial Internet of Things Manager
v003 IoT Software Engineer, v004 IoT Saas Architect
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 12 / 20
Context-aware Dynamic Text Field
Context-aware Dynamic Text Field (DTF) for input
recommendation 3
Improving collection of job seeker info via web survey
Methods Used
String distance
Co-occurrence probability of words
3 (S. Chala, F. Ansai, & M. Fathi, 2016)
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 13 / 20
Context-aware Dynamic Text Field
Added Values
Adaptive to context
Ergonomically suitable
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 14 / 20
Integrating Social Network Data
Enriching job seeker with social networking data 4
4 (S. Chala and M. Fathi, 2017)
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 15 / 20
Integrating Social Network Data
ID Reputation Description Up
vote
Down
vote
89 1173 Network,
Engineer,
Service,
Provider,
Networks,
IPv6,
Network,
Security
6 0
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 16 / 20
Integrating Social Network Data
ID Reputation Description Up
vote
Down
vote
89 1173 Network,
Engineer,
Service,
Provider,
Networks,
IPv6,
Network,
Security
6 0
network engineer service provider ipv6 security
3 1 1 1 1 1
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 16 / 20
Result
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 17 / 20
Conclusion
Unsupervised feature learning from vacancies and job seekers
Improved DTF user interface for data collection
Social networking data enhanced job seeker
Improves precision of matching
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 18 / 20
Future Work
Extract the required and desired skills when a vacancy has not
explicitly categorized them
Adding multilingual capability to support seamless labor force
mobility
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 19 / 20
Thank You
Thank You
Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 20 / 20

Sisay Chala

  • 1.
    Data-intensive Bidirectional Matchingof Job Seeker to Vacancy Sisay Adugna Chala Institute of Knowledge Based Systems and Knowledge Management University of Siegen sisay.chala@uni-siegen.de June 26, 2017 Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 1 / 20
  • 2.
    Sisay Adugna Chala(University of Siegen) EDUWORKS 2017 June 26, 2017 2 / 20
  • 3.
    Motivation Jobs are notfilled because of lack of the right applicant 1 Job seekers don’t have access to the right jobs Vacancies are not complete Applicants do not have complete information, e.g., preferences Jobs are not filled because of applicants quitting, being fired 2 Applicants under- or overstated their suitability to the job 1 (Belloni, M., Brugiavini, A., Meschi, E., and Tijdens, K. G., 2016) 2 (Branham, L., 2012) Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 3 / 20
  • 4.
    Motivation ... Huge onlinedata of job descriptions entered by job seekers and job holders Employers produce large volume of vacancy data online Due to this volume, not all vacancies are reachable by job-seekers who have relevant skill set Not all vacancies specify all the required skills Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 4 / 20
  • 5.
    Motivation ... Sisay AdugnaChala (University of Siegen) EDUWORKS 2017 June 26, 2017 5 / 20
  • 6.
    Matching Why is matchingof job seeker to vacancy is challenging as compared to other matching problems? Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 6 / 20
  • 7.
    Bidirectional Matching Definition Bidirectional Matchingmeasures the degree of semantic similarity of job seekers against vacancies and matches one to the other Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 7 / 20
  • 8.
    Methods and Tools Webmining – to scrap online vacancy and resumé data Natural language processing – to represent the textual data Machine learning – to extract, and present the analysis result Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 8 / 20
  • 9.
    Data Source Wageindicator dataset Onlinejob description data from occupational standards Online vacancy data via crawling Online resumé Social networking data Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 9 / 20
  • 10.
    Data Pre-processing Sisay AdugnaChala (University of Siegen) EDUWORKS 2017 June 26, 2017 10 / 20
  • 11.
    Document Similarity Analysisand Clustering Building a document vector in order to represent the document as a whole using statistically most important words contained in the document Similarity analysis applications work by comparing the vectors of documents using Latent Semantic Analysis (LSA) Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 11 / 20
  • 12.
    Baseline c000 Software DeveloperIntern c001 Business Development Manager IOT c002 Web Developer v000 Software Development and Integration Manager v001 Software Development Manager, IT integration v002 Industrial Internet of Things Manager v003 IoT Software Engineer, v004 IoT Saas Architect Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 12 / 20
  • 13.
    Baseline c000 Software DeveloperIntern c001 Business Development Manager IOT c002 Web Developer v000 Software Development and Integration Manager v001 Software Development Manager, IT integration v002 Industrial Internet of Things Manager v003 IoT Software Engineer, v004 IoT Saas Architect Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 12 / 20
  • 14.
    Context-aware Dynamic TextField Context-aware Dynamic Text Field (DTF) for input recommendation 3 Improving collection of job seeker info via web survey Methods Used String distance Co-occurrence probability of words 3 (S. Chala, F. Ansai, & M. Fathi, 2016) Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 13 / 20
  • 15.
    Context-aware Dynamic TextField Added Values Adaptive to context Ergonomically suitable Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 14 / 20
  • 16.
    Integrating Social NetworkData Enriching job seeker with social networking data 4 4 (S. Chala and M. Fathi, 2017) Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 15 / 20
  • 17.
    Integrating Social NetworkData ID Reputation Description Up vote Down vote 89 1173 Network, Engineer, Service, Provider, Networks, IPv6, Network, Security 6 0 Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 16 / 20
  • 18.
    Integrating Social NetworkData ID Reputation Description Up vote Down vote 89 1173 Network, Engineer, Service, Provider, Networks, IPv6, Network, Security 6 0 network engineer service provider ipv6 security 3 1 1 1 1 1 Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 16 / 20
  • 19.
    Result Sisay Adugna Chala(University of Siegen) EDUWORKS 2017 June 26, 2017 17 / 20
  • 20.
    Conclusion Unsupervised feature learningfrom vacancies and job seekers Improved DTF user interface for data collection Social networking data enhanced job seeker Improves precision of matching Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 18 / 20
  • 21.
    Future Work Extract therequired and desired skills when a vacancy has not explicitly categorized them Adding multilingual capability to support seamless labor force mobility Sisay Adugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 19 / 20
  • 22.
    Thank You Thank You SisayAdugna Chala (University of Siegen) EDUWORKS 2017 June 26, 2017 20 / 20