Building system for online job posting analysis using ESCO€¦ · •job search strategies By...
Transcript of Building system for online job posting analysis using ESCO€¦ · •job search strategies By...
25/10/ 2011
Building system for online job posting analysis using ESCO
Pascaline Descy, Vladimir KvetanDepartment for Skills and Labour Market
Cedefop
Emilio ColomboCRISP
Outline
• Description of the project
– (V. Kvetan)
• Using ESCO as the overarching taxonomy
– (E. Colombo)
• Relevance and wider context of the proejct
– (P. Descy)
Why do we need skills intelligence?
• Skills are key for Europe’s competitiveness
• Skills are key for economic growth
• Skills are key for individual’s chances
=> Understand current and future skill needs
=> Assess skills mismatch its causes and consequences
=> Share data and evidence with a wide audience
Why exploring online job postings?
No EU-wide data on skill demands by employers
Surveys
• expensive
• suffer from time lags
• focused on limited set of skills
Job postings can help
• Understand timely market demands
• Collect detailed data on job-specific and generic skills
• Broken down by detailed occupation and region
Feasibility of online job postings analysis
Multilingual web vacancy data in 5 countries
Final results• Regions• Occupations • Contract type and working hours• Skills and jobs requirements
– ESCO skills– Additional skills– Other job requirements– Experience
Feasibility study outcomes
Setting up a pan European instrument is feasible
ESCO can be used as the overarching taxonomy
Results will provide unique and new information
Blacksmiths and toolmakers What skills “make a difference”?
Painters in UK• Job specific skills
• Transversal skills
• New skills
Possible future data applications
Using ESCO
• ESCO is used both as input and as output
• Input: main vocabulary for classifying
– Occupations
– Skills
• Output: main framework for presenting results
ESCO V1 vs V0• ESCO V1 is a major improvement over V0
• It is richer and more detailed (particularly on skills)
• It contains synonyms and related words
• It is organised as a graph and not as a strict hierarchy
• It is multilanguage
Richness is important• In the prototype, the classification process required
experts constructing a training
• Using the training set a machine learning algorithm was able to classify all the vacancies
• Limitations: experts generally classify vacancies univocally. This narrows the knowledge base. Moreover small mistakes can be made.
– Example programmer. Expert 1 can classify it as Software developer (ISCO 2512). Expert 2 can classify as system analyst (2511). Correct at 3 digit, not at 4 digit
Richness is important• New system: enriched knowledge base with the
entire information content of ESCO.
– Example: the machine embeds in the learning algorithm allthe possible relations that ESCO contains for programmers
• Synonyms and related words are coded in originallanguage and not directly translated from English.Original language is similar to descriptions containedin vacancies
=> Knowledge base is richer and larger
Richness is important
• ESCO knowledge base forms the initial training set;experts validate it or work only on vacancies that aredifficult to classify
=> Using the same vacancies of the prototype wediscovered 50 additional occupations (ISCO 4 digit) incomparison to classifying with ESCO V0
Hierarchy vs graph• ESCO V0 was strictly hierarchical
• ESCO V1 is partially hierarchical and organised like a graph
Hierarchy vs graph
• A graph structure is better when dealing with totallyunstructured data such as vacancies since
• we need to interpret a piece of text that containswords that can be strongly or weakly related to theconcept we need to grasp
Setting up an wide EU system
2017-2018
• Landscaping EU online job market
• Extend coverage to more countries and websites
• Apply ESCO v1 to classify occupations and skills
• Release early data set for 7 countries (FR, ES, UK, DE, IT, CZ, IE)
2018-2020
• Cover all EU Member States
• Refine algorithms and classifiers
• Open data access
• Data integration into online EU tools (Europass, Skills Panorama, Eures)
Use of Data• Inform career mobility,
• CVT choices
• job search strategies
By job-seekers / career counsellors
• Fine-tune training offer
• Train for shortage occupations
By training providers / PES
• Refine recruitment and talent management strategiesBy employers
• Understand skills demand, emerging skills, skill gapsBy policy-makers
Education, training, lifelong learning
Employment and skills matching
Labour mobility
Detailed and market driven data for policy
Working together
Time is needed…
Burčiak (SK)Federweisser (DE)
Vin bourru (FR)Scrumpy (UK)
Víno (SK)Wein (DE)Vin (FR)
Wine (UK)