9/7/2017 © 2017 CareerBuilder
Learn From the Experts: How CareerBuilder Uses
Innovation to Solve Talent Acquisition Problems
ML/AI Edition
9/7/2017 © 2017 CareerBuilder
#CBEmpower
Learn from the Experts
© 2017 CareerBuilder2
Introduction
3
As VP of Global Services Strategy, Matt
focuses on the edges of CareerBuilder’s
products, ensuring that they drive recruiter
efficiency by integrating well with each
other and with the tools recruiters use daily.
His 12 years of experience in the
recruitment space has made him passionate
about applying data science, running high
scale microservices, and bringing it all
together into a recruiter friendly sourcing
product.
Matt McNair
CareerBuilder
Introduction
4
Rob Houser, Senior Director of UX, has
worked in the field of user experience for
25 years with companies such as NCR
Corporation (Retail Solutions) and Sage
Software (Accounting/ERP, Payroll, and
HCRM software). Rob manages a team of
user researchers, interaction designers, and
visual designers who work on
CareerBuilder’s software products.
Rob Houser
CareerBuilder
What is Automation?
5
Automation is the use of various
control systems to move manual
operations or those done by
humans to processes that require
minimal or reduced human
intervention.
What is Machine Learning?
6
Machine Learning is a current
application of AI based around the
idea that we should really just be
able to give machines access to
data and let them learn for
themselves.
What is Artificial Intelligence?
7
Artificial Intelligence is the broader
concept of machines being able to
carry out tasks in a way that we
would consider “smart”.
What is a Knowledge Graph?
8
The Knowledge Graph is
a knowledge base used by Google to
enhance its search engine's search
results with semantic-search
information gathered from a wide
variety of sources.
What is a Deep Learning?
9
Deep learning (also known as deep
structured learning or hierarchical learning)
is the application of artificial neural
networks (ANNs) to learning tasks that
contain more than one hidden layer. Deep
learning is part of a broader family
of machine learning methods based
on learning data representations, as
opposed to task-specific algorithms.
What is a Unified Taxonomy?
10
Taxonomic data form a substantial, but
scattered, resource. The alternative to such
a fragmented system is a 'unitary' one of
preferred, consensual classifications. For
effective access and distribution the (Web)
revision for a given taxon would be
established at a single Internet site.
Classification of Job Seekers | Browse Concept
© 2017 CareerBuilder11
What Self Driving Cars See
© 2017 CareerBuilder12
User Experience Should Drive Good Design
© 2017 CareerBuilder13
Using ML/AI To Intelligently Suggest Candidates
© 2017 CareerBuilder14
Searches I Ran Yesterday
Using ML/AI To Intelligently Suggest Candidates
© 2017 CareerBuilder15
From All My Sources
Semantic Search in RDB
© 2017 CareerBuilder16
9/7/2017 © 2017 CareerBuilder
#CBEmpower
Research and Development
Introduction
18
Yun Zhu is the Lead Data Scientist of Applied
Machine Learning team at CareerBuilder.
His team builds the underlying machine
learning models that CareerBuilder
products rely on.
Name
Company Name
Yun Zhu
CareerBuilder
Introduction
19
Khalifeh Al Jadda is the lead data scientist
on the Search Data Science Team. He has
experience implementing large scale,
distributed machine learning algorithms to
solve challenging problems in domains
ranging from Bioinformatics to search and
recommendation engines. He leads the data
science projects to improve search
relevancy and build a new recommendation
engine leveraging AI, Machine Learning, and
Big Data frameworks.
Name
Company Name
Khalifa AlJadda
CareerBuilder
Context Around Skills | Search for Sales Professionals
© 2017 CareerBuilder20
Helpful
Less Helpful
Observed Signals Inferred Signals
Likelihood To Respond
© 2017 CareerBuilder21
Candidate Tenure
Is Beyond The
Expected Tenure
For This RoleCandidate Active
Yesterday
Will Robots Take My Job?
22 http://www.bbc.com/news/technology-34066941
We’re Safe.
© 2017 CareerBuilder23
9/7/2017 © 2017 CareerBuilder
#CBEmpower
Semantic Search
Introduction
25
Founder and CEO of Textkernel (2001). Since 2015
part of CareerBuilder group. Leads team of 110
people, most tech experts. R&D background in AI,
Machine Learning & NLP since 1990.
Born in Prague, lives in Amsterdam, married, 2
teenage daughters. Eclectic music tastes.
Semantic Recruitment / Labour Market Analysis /
Natural Language Processing / Machine Learning /
Search & Match / Enterprise Software
@jakubzavrel zavrel@textkernel.nl
Jakub Zavrel
Textkernel
What is Semantic Search?
Find what
you mean
...
not what
you type
Search for
things
...
not for
strings
Semantic Search & Match. Why?
More relevant candidates, faster!
1. More candidates: Expanding words to concepts
No guessing what candidate wrote in the resume
2. Relevant candidates: Better filtering and ranking
Avoid wrong matches by understanding context
3. Work faster
- Start working on the most promising candidates first
- Be a great searcher without advanced Boolean search
- Use documents (jobs, resumes) to start automated searches
A twoword user query
= 35+ system query
3 Elements of Semantic Search
Document
understanding
Domain
knowledge
Machine
Learning
➔CV & job parsing
➔Information
extraction
➔Taxonomies
➔Ontologies
➔Skills
➔Keywords
➔Algorithms learn
from a large set
of data
➔Learning to Rank
➔Deep Learning
The job is
automatically
converted into a
search query..
.. and matched to the
candidates in your
database(s)
... and vice versa!
Semantic Expansion Example
9/7/2017 © 2017 CareerBuilder
Thank you
 Scott.helmes@careerbuilder.com  Twitter: @shelmes
How did we do? You can tell us
now by visiting the app to provide
your review.

How CareerBuilder Uses Innovation to Solve Talent Acquisition Problems

  • 1.
    9/7/2017 © 2017CareerBuilder Learn From the Experts: How CareerBuilder Uses Innovation to Solve Talent Acquisition Problems ML/AI Edition
  • 2.
    9/7/2017 © 2017CareerBuilder #CBEmpower Learn from the Experts © 2017 CareerBuilder2
  • 3.
    Introduction 3 As VP ofGlobal Services Strategy, Matt focuses on the edges of CareerBuilder’s products, ensuring that they drive recruiter efficiency by integrating well with each other and with the tools recruiters use daily. His 12 years of experience in the recruitment space has made him passionate about applying data science, running high scale microservices, and bringing it all together into a recruiter friendly sourcing product. Matt McNair CareerBuilder
  • 4.
    Introduction 4 Rob Houser, SeniorDirector of UX, has worked in the field of user experience for 25 years with companies such as NCR Corporation (Retail Solutions) and Sage Software (Accounting/ERP, Payroll, and HCRM software). Rob manages a team of user researchers, interaction designers, and visual designers who work on CareerBuilder’s software products. Rob Houser CareerBuilder
  • 5.
    What is Automation? 5 Automationis the use of various control systems to move manual operations or those done by humans to processes that require minimal or reduced human intervention.
  • 6.
    What is MachineLearning? 6 Machine Learning is a current application of AI based around the idea that we should really just be able to give machines access to data and let them learn for themselves.
  • 7.
    What is ArtificialIntelligence? 7 Artificial Intelligence is the broader concept of machines being able to carry out tasks in a way that we would consider “smart”.
  • 8.
    What is aKnowledge Graph? 8 The Knowledge Graph is a knowledge base used by Google to enhance its search engine's search results with semantic-search information gathered from a wide variety of sources.
  • 9.
    What is aDeep Learning? 9 Deep learning (also known as deep structured learning or hierarchical learning) is the application of artificial neural networks (ANNs) to learning tasks that contain more than one hidden layer. Deep learning is part of a broader family of machine learning methods based on learning data representations, as opposed to task-specific algorithms.
  • 10.
    What is aUnified Taxonomy? 10 Taxonomic data form a substantial, but scattered, resource. The alternative to such a fragmented system is a 'unitary' one of preferred, consensual classifications. For effective access and distribution the (Web) revision for a given taxon would be established at a single Internet site.
  • 11.
    Classification of JobSeekers | Browse Concept © 2017 CareerBuilder11
  • 12.
    What Self DrivingCars See © 2017 CareerBuilder12
  • 13.
    User Experience ShouldDrive Good Design © 2017 CareerBuilder13
  • 14.
    Using ML/AI ToIntelligently Suggest Candidates © 2017 CareerBuilder14 Searches I Ran Yesterday
  • 15.
    Using ML/AI ToIntelligently Suggest Candidates © 2017 CareerBuilder15 From All My Sources
  • 16.
    Semantic Search inRDB © 2017 CareerBuilder16
  • 17.
    9/7/2017 © 2017CareerBuilder #CBEmpower Research and Development
  • 18.
    Introduction 18 Yun Zhu isthe Lead Data Scientist of Applied Machine Learning team at CareerBuilder. His team builds the underlying machine learning models that CareerBuilder products rely on. Name Company Name Yun Zhu CareerBuilder
  • 19.
    Introduction 19 Khalifeh Al Jaddais the lead data scientist on the Search Data Science Team. He has experience implementing large scale, distributed machine learning algorithms to solve challenging problems in domains ranging from Bioinformatics to search and recommendation engines. He leads the data science projects to improve search relevancy and build a new recommendation engine leveraging AI, Machine Learning, and Big Data frameworks. Name Company Name Khalifa AlJadda CareerBuilder
  • 20.
    Context Around Skills| Search for Sales Professionals © 2017 CareerBuilder20 Helpful Less Helpful
  • 21.
    Observed Signals InferredSignals Likelihood To Respond © 2017 CareerBuilder21 Candidate Tenure Is Beyond The Expected Tenure For This RoleCandidate Active Yesterday
  • 22.
    Will Robots TakeMy Job? 22 http://www.bbc.com/news/technology-34066941
  • 23.
    We’re Safe. © 2017CareerBuilder23
  • 24.
    9/7/2017 © 2017CareerBuilder #CBEmpower Semantic Search
  • 25.
    Introduction 25 Founder and CEOof Textkernel (2001). Since 2015 part of CareerBuilder group. Leads team of 110 people, most tech experts. R&D background in AI, Machine Learning & NLP since 1990. Born in Prague, lives in Amsterdam, married, 2 teenage daughters. Eclectic music tastes. Semantic Recruitment / Labour Market Analysis / Natural Language Processing / Machine Learning / Search & Match / Enterprise Software @jakubzavrel zavrel@textkernel.nl Jakub Zavrel Textkernel
  • 26.
    What is SemanticSearch? Find what you mean ... not what you type Search for things ... not for strings
  • 27.
    Semantic Search &Match. Why? More relevant candidates, faster! 1. More candidates: Expanding words to concepts No guessing what candidate wrote in the resume 2. Relevant candidates: Better filtering and ranking Avoid wrong matches by understanding context 3. Work faster - Start working on the most promising candidates first - Be a great searcher without advanced Boolean search - Use documents (jobs, resumes) to start automated searches A twoword user query = 35+ system query
  • 28.
    3 Elements ofSemantic Search Document understanding Domain knowledge Machine Learning ➔CV & job parsing ➔Information extraction ➔Taxonomies ➔Ontologies ➔Skills ➔Keywords ➔Algorithms learn from a large set of data ➔Learning to Rank ➔Deep Learning
  • 29.
    The job is automatically convertedinto a search query.. .. and matched to the candidates in your database(s) ... and vice versa! Semantic Expansion Example
  • 30.
    9/7/2017 © 2017CareerBuilder Thank you  Scott.helmes@careerbuilder.com  Twitter: @shelmes How did we do? You can tell us now by visiting the app to provide your review.