Loading presentation...

Present Remotely

Send the link below via email or IM


Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.


Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

Textkernel SOSU Europe 13 Sept 2013

No description

Jakub Zavrel

on 15 April 2014

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of Textkernel SOSU Europe 13 Sept 2013

Thank you!
Post and pray model is broken
ATS candidate databases
Social networks
Market transparency
So why is not everyone doing this?
Semantic Search
Active Sourcing & Semantic Search
Sourcing Summit Europe, 12-13 September Amsterdam
Jakub Zavrel

Semantic recruitment technology - the solution?
What is "Semantic Search"?
A search engine that:

lets you find what you mean,
...not what you type.
what your
query means
what your
search results
8 Elements of
1. Understanding what you type

Robustness against typo's
Query enhanced with
spelling variants
2. Disambiguation
same word, different context

Meaning in context
Know which meaning more likely
Fielded search
tagclouds, etc.
3. Concepts not keywords

Broader and narrower terms
Related terms
4. Similarity and relatedness

Always get best possible results
Meaningful relevance ranking
Search for similar
Identify clusters
5. Parts and wholes

Specific meaning of separate query parts
Impact of each part on the relevance
Combining multiple criteria
7. Fully natural language search
No query syntax to learn, no complex advanced search forms
Search in your own language
Job description as a query
6. Relationships
Reasoning about relationships
Fusing information
Predictive models
8. Meaningful results

Understandable and actionable
Answers vs. documents
Merging / fusion of objects
Across language boundaries
Textkernel Semantic Search & Match
1. We're not there yet...
2. As active sourcing becomes accessible to mainstream recruiters, we will finally be able to move past the "Post and Pray" model
3. Companies with better sourcing knowledge will be better competitors in the market
4. Cost per hire is going down
5. The labor market is really going to become more transparent
7. To predict is the future of sourcing...
6. Social, marketing and selling skills will become the more important part of the recruitment skillset
8. No Boolean Black Magic needed to be a great recruiter
MSc Artificial Intelligence
R&D in Language Technology
CEO Textkernel
1. Understanding what you type
2. Same word different context
3. Concepts not keywords
4. Similarity and relatedness
5. Parts and wholes
6. Relationships
7. Fully Natural Language Search
8. Meaningful results
Active Sourcing
Recruiters have difficult queries:
Find me a parttime legal secretary in Utrecht (English, Dutch, French) for small law office

Looking a for an experienced German speaking front-end developer with knowledge of Java, jQuery, MySQL, and CSS

Pipeline engineer offshore, North Sea, Rig up/ down equipment unsupervised, HSE

Manager Biogenic Materials, Eindhoven, French speaking, pricing and trading, HBO level

Sales Manager, HR Technology, Frankfurt am Main Area, Software experience, good network & marketing skills
And what if it also works for the mainstream recruiter...?
Post & Pray

Structured search: No data entry or complex queries
Easy visual interface
Strong ranking of results vs. filtering
One dashboard for internal and external search
Knowledge of recruitment language in the box
Natural Language Query Understanding
Query with a Job ad
Multi-lingual synonyms in the box
Search multiple social networks and CV databases
Set up search alerts / manage saved results
Standalone or integrate into your ATS

Made for active sourcing!

Ask me for a demo!
The brave new world of semantic sourcing
e.g. www.booleanblackbelt.com
Structured queries on unstructured data
Dealing with language variations:
Spelling, Abbreviations, Synonyms
Context: same word can mean two (or more) things
Concepts and relationships, not keywords
Multiple languages
Boolean Black Magic is a Poor Man's Semantic Search
Make Recruitment Technology Meaningful
Boolean Search
- spelling
- synonyms
- related concepts
- multi-lingual variation
- geographical ranges

- to join more than one criterion
- filtering based on context

- eliminate unwanted results



Current * (developer OR programmer)
(Siemens OR Cisco OR Avaya)
on time
Boolean Search:
returns the set of results that satisfy a logical formula of expressions (keywords) tied together with OR, AND, NOT operators...

- Very precise definition of a set
- Complex expressions possible
- Clear logical model

- Habit:
- We've gotten used to it
- Pretty standard in most search engines

- Good for your social status
- Wrong model?
- Are we really looking for a strict set?
- No ranking within the set

- Expressions are complex to debug
- Hard to balance recall and precision

- Hard to master for non-technical people
- Bad for broad adoption of sourcing
Precision: % of good candidates as part
of all results returned.

Recall: % of all relevant candidates in
my results
Boolean Black Magic is a Poor Man's Semantic Search
That's not:
But more like:
Full transcript