DLV011 – Summary report on ESCO skills pillar pilot
1. The current structure of the ESCO skills pillar ........................................... 6
2. Current trends ............................................................................................ 12
3. Objective and scope of the pilot ................................................................ 14
4. The methodology of the pilot .................................................................... 14
5. Results and lessons learnt .......................................................................... 19
6. Recommendations ..................................................................................... 26
7. Open issues ................................................................................................ 27
DLV011 – Summary report on ESCO skills pillar
pilot
Document version and date : Version 1.1
2018-05-16
DLV011 – Summary report on ESCO skills pillar pilot
2 of 28
Document Metadata Property Value
Release date 2018-05-16
Status Draft
Version 1.0
Authors Aikaterini Sylla
Honza Förster
Reviewed by
Approved by
Document History Version Date Description Action
1.0 2018-04-30 First release For acceptance
1.1 2018-05-16 Updated release For acceptance
DLV011 – Summary report on ESCO skills pillar pilot
3 of 28
Executive summary
Purpose of the document
The purpose of this document is to provide the ESCO Maintenance Committee with a
short overview of the ESCO skills hierarchy1 pilot which was carried out between 10
November 2017 and 15 March 2018. This would allow the ESCO Maintenance
Committee to become more familiar with the ESCO skills hierarchy in view of the June
2018 meeting, where the Commission intends to discuss with the committee the further
development of the skills pillar.
The scope of the pilot
In the context of the skills hierarchy pilot, the ESCO team tested various approaches to
structure further the ESCO skills and assess how well those approaches can support the
user needs described above. Those needs are reflected in examples of use cases, as
follows:
Find a concept in the classification
Select groups of concepts for a specific action
The system suggests related concepts
Within the framework of this pilot, the ESCO team tested several structuring approaches
on a sample of the ESCO skills pillar. These are the following:
Skills hierarchy:
DISCO (European Dictionary of Skills and Competences)2
Kompetenzklassifikation (the skills classification used by the German PES)3
Grouping skills by using other domains as structure/collections:
NACE (Statistical Classification of Economic Activities in the European Union);
ISCED-F (International Standard Classification of Education - Fields of
Education and Training);
CPA (Statistical classification of products by activity);
work context using the work "Arbeitsorte" in the Kompetenzklassifikation
Contextualisation:
analysis of how transversal skills, competences or knowledge are applied in the specific
context of a sector or an occupation.
1 Pyramid-like ranking of skill where every level (except the top and the bottom ones) has one
higher and one lower neighbour. Lower level means more granularity. 2 http://discotools.eu/disco2_portal/ 3 https://download-portal.arbeitsagentur.de/files/login.do
DLV011 – Summary report on ESCO skills pillar pilot
4 of 28
Results and lessons learnt
Skills hierarchy
The results of structuring skills with the hierarchy of Kompetenzklassifikation and
DISCO are similar since both classifications i) follow a similar approach to structuring
skill concepts, ii) use a hierarchical grouping of concepts and iii) use sectors of economic
activity to structure sector-specific skills.
Neither contains detailed descriptions of their skill groups that could clarify their scope,
which makes it difficult to assign ESCO skills to the respective groups. Furthermore, both
classifications use economic sectors to classify domain-specific skills. Due to this, they
cannot be easily used as a mono-hierarchical structure, where each skill fits exactly in one
skill group.
Using other domains as structure/collection
Another approach to structuring skills is by expressing how they relate to other types of
concepts. This can be particularly useful for filtered search. The existing relationship
between skills and occupations in ESCO is of this type: it expresses which skills are
normally relevant in the context of an occupation and thus allows the user to filter the
skills pillar by occupation.
NACE
This would allow users to filter on skills that typically contribute to a specific
economic activity. “Load animals for transportation” could, for example, be a skill
relevant, among others, in the economic activities “freight transport by road” and
“animal production”.
In the pilot, several transversal skills were assigned to all NACE sections since they
apply to all sectors of economic activity. There was hardly any ESCO skill that could
be related to only one single NACE section.
The definitions of the NACE sections and the indications of the scope of the section
are very useful to delimit the scope and be accurate in assigning skills.
ISCED-F
The ESCO team assigned skills to the domains of education and training, where these
skills are usually taught. “Load animals for transportation” could, for example, be a
skill taught in education and training programs in the field of “transport services” or
“crop and livestock production”.
The results showed that the ISCED-F classification is valuable and would be
particularly useful when annotating learning outcome descriptions.
CPA
The CPA classification lists services and products in 3,218 subcategories. It proved
difficult to decide on the appropriate relation to skills. It could, for example, express
that knowledge of a certain product or service category is useful for developing a
skill. For example, knowledge of “road transport services of live animals” or of “dairy
cattle, live” is useful for developing the skill “load animals for transportation”.
DLV011 – Summary report on ESCO skills pillar pilot
5 of 28
Work Context
The Kompetenzklassifikation work context classification is limited to 156 concepts
which are a small number to assign to 13,485 ESCO skills, whereas it is usually not
possible to assign a skill to a single work context. Skills are usually performed in a
range of different locations.
Contextualisation
Contextualisation relations can be established if the description of the skills is clear
enough to define its content and scope. The approach was used when developing ESCO
v1, but it has not been applied consistently over the entire skills pillar.
Therefore, the ESCO team proposes a set of rules for improving the skills descriptions.
This will allow creating more precise and consistent relations. The rules will be listed in
the final report of the pilot project.
Recommendations
Following the results of the pilot, the ESCO team recommends further exploring the use
of ISCED-F and NACE as external classifications to structure ESCO.
Contextualisation is a useful methodology since it provides clear relations among skill
concepts. It can be used for various purposes and the skills pillar is already annotated
with such relations. Furthermore, revising the contextualisation and the methodological
rules related to it would improve the consistency of the skill pillar. On the other hand, it
requires a high effort to manually maintain such relations due to a high number of
concepts in the skills pillar.
DISCO and Kompetenzklassifikation, the two hierarchical skills classifications
evaluated within the scope of this project, use sectors of economic activity to structure
skills. Sectors of economic activity are not suitable as a base for a hierarchical
arrangement of concepts in which each concept can have only one broader concept
(mono-hierarchy), which is the hierarchical arrangement of interest in the case of ESCO,
because a single skill can inherently be related to several different sectors of economic
activity.
Pending issues
The main pending issue identified during the pilot is the lack of an existing suitable
mono-hierarchy that could be used for statistical labour market data purposes. This could
be the default browsing mechanism of the skills pillar. Such classification would have to
be composed of skill groups which are mutually exclusive, i.e. a single skill can be placed
only in a single group, thanks to clear scoping of groups. For the ESCO occupational
pillar, ISCO serves as such a mono-hierarchical structure.
In case a mono-hierarchy of skills would be the most suitable structure to address the user
needs which are reflected in the use case examples, it would be necessary to build a
custom solution for the purposes of ESCO.
DLV011 – Summary report on ESCO skills pillar pilot
6 of 28
Summary report on the ESCO skills pillar pilot
1. The current structure of the ESCO skills pillar
The skills pillar of ESCO v1 structures 13,485 concepts in four different manners:
• through their relationship with occupations, by using occupations as an entry
point;
• through a hierarchy (only for transversal knowledge, skills and competences);
• through associative4 and contextualised
5 relations (not consistently applied
throughout the entire classification);
• through functional collections grouping only a subset of the skills available that
is of interest for examples of use cases. ESCO v1 includes three functional
collections: digital competences (identical to the Digital Competence
Framework6); language skills and transversal skills.
Covering certain user needs requires the improvement of ESCO's current structure. Such
user needs would include searching for specific skills, doing semantic search, filtering
search results, clustering skills in groups (e.g. language skills, digital skills or
management skills) to work at a more aggregated level, identifying related concepts,
identifying concepts of a specific interest for the user, using only part-subset of the
classification, getting statistics etc. To address this need, the ESCO team developed and
carried out a pilot to arrive at a structure of the ESCO skills pillar that would better serve
such functionalities.
Each one of the manners is described in the sections that follow.
1.1. Relationship with occupations
Each ESCO occupation is related to essential and optional knowledge, skill and
competence concepts. The image on the next two pages shows an example of the
information attached to each ESCO occupation and its relationship to knowledge, skills
and competences.
4 Associative relations indicate how knowledge, skills and competences are relevant to other knowledge, skills and
competences. 5 Skill contextualisation is a method to create knowledge or skill and competence concepts by analysing how
transversal skills, competences or knowledge are applied in the specific context of a sector or an occupation. For more
information and example, you can visit the ESCO portal:
https://ec.europa.eu/esco/portal/escopedia/Skill_contextualisation
6 https://ec.europa.eu/jrc/en/digcomp
DLV011 – Summary report on ESCO skills pillar pilot
7 of 28
DLV011 – Summary report on ESCO skills pillar pilot
8 of 28
7
Figure 1: ESCO Occupation profile
1.2. Hierarchical relations in the transversal skills
ESCO provides metadata for each concept in the skills pillar. Part of this is the skill
reusability level, which indicates how widely a knowledge, skill or competence concept
can be applied. This is crucial for supporting occupational mobility. ESCO distinguishes
four levels of skill reusability:
• Transversal knowledge, skills and competences: they are relevant to a broad
range of occupations and sectors;
• Cross-sector knowledge, skills and competences: they are relevant to
occupations across several economic sectors;
7 ESCO Handbook, p.15-16. Available at: https://ec.europa.eu/esco/portal/document/en/0a89839c-098d-4e34-846c-
54cbd5684d24
DLV011 – Summary report on ESCO skills pillar pilot
9 of 28
• Sector-specific knowledge, skills and competences: they are specific to one
sector, but are relevant for more than one occupation within that sector;
• Occupation-specific knowledge, skills and competences: they are usually
applied only within one occupation or specialism.
Transversal knowledge, skills and competences are relevant to a broad range of
occupations and economic sectors. They are often referred to as core, basic or soft skills
and are the cornerstone of the personal development of a person.
Within the skills pillar, transversal skills and competences are organised in a hierarchical
structure with the following five headings:
• thinking
• language
• application of knowledge
• social interaction
• attitudes and values
Both the concepts and hierarchical structure of the transversal knowledge, skills and
competences were developed by the Cross-sector Reference Group.8 The development
was based on the analysis of a wide range of existing national and sectoral classifications,
the European Dictionary of Skills and Competences (DISCO)9 and other sources.10 In the
transversal skills, there are hierarchical relations.
Example: The skill group “social interaction” contains as narrower concepts the
following skills, among others:
• report facts;
• persuade others;
• give advice to others.
In addition, there are some hierarchical relations between two skills or small sets which
are not linked to the transversal skills.
Example: The skill “analyse ICT system” is a sector-specific skill which is broader to the
sector-specific skill “identify ICT system weaknesses”.
1.3. Contextualised knowledge, skills and competences
Skill contextualisation is a method to create knowledge or skill and competence concepts
by analysing how transversal skills, competences or knowledge are applied in the specific
context of a sector or an occupation. This allows transversal knowledge, skills and
competences that are abstract to be brought to a more detailed level so that they can be
directly used in occupational profiles.
8 For information about the Cross-sector Reference Group see ESCO Handbook, p.38. Available at:
https://ec.europa.eu/esco/portal/document/en/0a89839c-098d-4e34-846c-54cbd5684d24
9 http://disco-tools.eu/disco2_portal/
10 ESCO Handbook, p.19-20. Available at: https://ec.europa.eu/esco/portal/document/en/0a89839c-098d-4e34-846c-
54cbd5684d24
DLV011 – Summary report on ESCO skills pillar pilot
10 of 28
Example: The skill “measure” is too abstract to be linked directly to the occupation
“metal furnace operator”. This relationship would produce too many results if used in
competence-based job matching since measuring is relevant for a large number of
occupations and sectors. Through skill contextualisation, the skill can be made more
specific. A skill named “measure furnace temperature” could, for example, be used in the
occupational profile of the “metal furnace operator”.11
Example: The skill “report facts” is too abstract to be linked directly to any sector.
Through skill contextualisation, the skill can be made more specific:
skill named “report pollution incidents” could, for example, be used in the sector of
“agriculture”;
skill named “report mine machinery repairs” could be used for the mining industry;
skill named “report anomalies in aircraft interiors” could be used in the sector of
transport.
1.4. Associative relations
Apart from hierarchical relations, there are also associative relations between skills. An
associative relation between two skills indicates that a skill is essential or optional to
carry out another skill.
Example: The skill “think creatively” is essential to design greeting cards and write
storylines.
11 ESCO Handbook, p.20. Available at: https://ec.europa.eu/esco/portal/document/en/0a89839c-098d-4e34-846c-
54cbd5684d24
DLV011 – Summary report on ESCO skills pillar pilot
11 of 28
The diagram below illustrates the current hierarchical and associative relations in the
ESCO skills pillar. The green arrows refer to the hierarchical relations while the red
arrows to the associative relations. The diagram can be interpreted as follows:
• the triangle depicts the hierarchical relations among the transversal skills;
• the skills in the red box are contextualized ones. The green arrows show the
relations among them but also with the transversal skills in the triangle;
• the rest of the relations on the left reflect relations among small sets of skills
which are not linked to the transversal skills;
• the skills with no relation established are orphans.
Figure 2: ESCO Associative relations
1.5. Functional collections
Functional collections allow subsections of the skills pillar to be selected, according to
the purpose it is going to be used for.
Example: An organisation may want to use ESCO to implement an online CV editor
where a user can indicate his/her language skills. The organisation would not need all
the ESCO skills in that CV section, only the language skills. If a user searches for
"Chinese" in this section, the system should suggest "Chinese", "understand spoken
Chinese", “understand written Chinese” or "interact verbally in Chinese", but not
"traditional Chinese medicine" or "give shiatsu massages".
A functional collection would allow the user to pick exactly the skills (or occupations)
s/he is looking for. ESCO v1 includes three functional collections: Digital transversal
skills (identical to the Digital Competence Framework); Language skills; Transversal
skills.12
12 ESCO Handbook, p.20-21. Available at: https://ec.europa.eu/esco/portal/document/en/0a89839c-098d-4e34-846c-
54cbd5684d24
DLV011 – Summary report on ESCO skills pillar pilot
12 of 28
2. Current trends
This section describes external ongoing initiatives which deal with the structuring of skill
information and labour market data in general:
2.1. Google Cloud Job Discovery (formerly Cloud Jobs API)
Google Cloud Job Discovery13 is a service which enables third party services, e.g. job
boards, to enrich their search algorithm with Google data based on machine learning
analysis of the labour market. Google defines the tools as follows:
“Cloud Job Discovery is part of Google for Jobs - a Google-wide commitment to help
people find jobs more easily. Job Discovery provides plug and play access to Google’s
search and machine learning capabilities, enabling the entire recruiting ecosystem -
company career sites, job boards, applicant tracking systems, and staffing agencies to
improve job site engagement and candidate conversion.”
The tool allows third-party services to perform a wide range of enrichment, such as:
- Recognise jargon used by employers in job vacancies and by job searchers:
Figure 3: Recognising jargon
13 https://cloud.google.com/job-discovery/
DLV011 – Summary report on ESCO skills pillar pilot
13 of 28
- Recognise searches which have multiple interpretations and variations in order to
list the most relevant results:
Figure 4: Understanding underlying concept
2.2. Cedefop Job vacancy search
Job vacancy search is an initiative, currently under development, by Cedefop to carry out
analysis of vacancies and emerging skill requirements across all EU Member States. The
real-time data collection system is aimed at collecting relevant background characteristics
of jobs, firms and the type of employee they are looking for, e.g. skills, qualifications and
other attributes, to enable the analysis of skills demand.
Once it is completed, Cedefop will be able to retrieve job vacancy information from job
boards of Member states and provide analysis on this data. The knowledge gathered can
then be used, among many other uses, to improve the ESCO classification by providing
real-world information on the usage of skills and occupations. Such statistics can drive
the development of content and terminology of ESCO during its continuous
improvement.
2.3. OPENSKIMR
Open European Skill Match Maker (OPENSKIMR) is an EU-funded project that aims to
create a skill-matchmaking between talents, and jobs and the required learning to support
people in creating their personal career routes. This project is based on:
ESCO
a set of algorithms that match talents with the perfect fitting job and in addition,
suggest learnings which shall be supported in evolving personal career routes of
future STEM talents.
It is a tool which aims to support people individually with their personal career path and
combines life-long-learning opportunities with specific job offers.
DLV011 – Summary report on ESCO skills pillar pilot
14 of 28
3. Objective and scope of the pilot
In the context of the skills hierarchy pilot, the ESCO team tested various approaches to
structure further the ESCO skills and assess how well those approaches can support the
user needs described above. Those needs are reflected in examples of use cases, as
follows:
• Find a concept in the classification
Example: an education provider searches for skills to insert into the learning outcomes
description of a qualification
• Select groups of concepts for a specific action
Example: an employer wants to select a group of skills when preparing a job vacancy
description, instead of picking all skills individually
• The system suggests related concepts
Example: A candidate receives suggestions for skills to complement her/his CV.
4. The methodology of the pilot
4.1. Select the ESCO skill dataset sample
The subset should include cross-sectoral, sector and occupation-specific skills as well as
the entirety of the transversal skills. The selection was based on ISCO-08 using the
relationship between skills and occupation profiles. As a result, the ESCO team extracted
in total 610 skills from the ESCO occupations related to the following two major ISCO
groups:
• ISCO-08 514: Hairdressers, beauticians and related workers
• ISCO-08 2432: Public relations professionals
These include:
skills/competences that are linked to the ESCO occupations of each of the ISCO
groups
skills/competences that are linked to one or more occupations of both ISCO
groups
skills/competences that are transversal to occupations of more ISCO groups
ISCO group ESCO occupations ESCO skills
514: Hairdressers, beauticians
and related workers 17 occupations 272
2432: Public relations
professionals 11 occupations 218
514 and 2432 13 cross-sectoral skills
514 and 2432 + more 107 transversal skills
Table 1: Overview of skill sample
DLV011 – Summary report on ESCO skills pillar pilot
15 of 28
4.2. Select the structuring approaches
The objective of the pilot was to explore various approaches to further structure the
ESCO skills and assess how well these approaches can support the user needs described
in section 3 Objective and scope of the pilot.
Within the framework of this pilot, the ESCO team tested several structuring approaches
on a sample of the ESCO skills pillar. These are the following:
• Skills hierarchy:
o Applying the hierarchy of DISCO (European Dictionary of Skills and
Competences) to ESCO skills14
;
o Applying the hierarchy of Kompetenzklassifikation (the skills
classification used by the German PES) to ESCO skills15;
• Grouping skills by using other domains as structure/collections:
o By economic activity using NACE (Statistical Classification of Economic
Activities in the European Union)16
;
o By fields of education and training using ISCED-F (International
Standard Classification of Education - Fields of Education and Training)
o By product domain using CPA (Statistical classification of products by
activity)
o By work context using the work location ("Arbeitsorte") in the
Kompetenzklassifikation
• Contextualisation: analysis of how transversal skills, competences or knowledge
are applied in the specific context of a sector or an occupation.
4.3. Assign the ESCO skills to each approach
The ESCO team assigned the skills to NACE, DISCO and Kompetenzklassifikation
manually using spreadsheet dataset per classification:
a spreadsheet listing the ESCO skills sample and
a spreadsheet listing metadata of the external classifications.
Using the information in these files the team assigned each ESCO skill from the sample
to a related concept (or multiple concepts in case of NACE) in the external
classifications. The relations were established using machine-readable fields, e.g. concept
IDs) to enable loading into the testing application.
With respect to the approaches of grouping skills using other domains as
structures/collections and the contextualisation, a different methodology was used.
14 DISCO is a structured vocabulary for the description of skills and competences in different contexts. Its development
was co-financed by the Commission. The top level of DISCO is divided into domain-specific skills and non-domain
specific skills. http://disco-tools.eu/disco2_portal/
15 Kompetenzklassifikation lists occupations and professional activities in various labour market and socio-economic
contexts.
https://berufenet.arbeitsagentur.de/berufenet/faces/index;BERUFENETJSESSIONID=4EZ16arPZ9VHK8807Z7DihYjl
2Uv_swUl5TI619mlkATWaxzAITQ!-132602081?path=null
16 NACE classifies economic activities in the EU. It is organised in 21 sections and indicates 4-digit codes common to
all European countries. http://ec.europa.eu/eurostat/statistics-
explained/index.php/Glossary:Statistical_classification_of_economic_activities_in_the_European_Community_(NACE
)
DLV011 – Summary report on ESCO skills pillar pilot
16 of 28
This is described in sections 5.2 Other domains as the structure and 5.3
Contextualisation.
4.4. Test the approaches
In order to test how the different structuring approaches, perform in real-life situations,
the ESCO team analysed each one of them on a set of predefined use-cases via a web-
based testing application. These are the following:
• Find a concept in the classification
• Select groups of concepts for a specific action
• The system suggests related concepts
4.4.1. Testing application
The project team developed a test application to test the skills structures.
The application displays the following two screens:
1. Search screen:
Figure 5: a Search screen
The search screen allows the user to look for a specific skill from the sample using a
search engine [1] identical to the one on the ESCO portal. The results are listed under [2]
“Text match”.
The user can further narrow down these results using filters [3] which are populated by
three of the structuring approaches we are testing, i.e. NACE, DISCO and NSC (National
Skills Classification17). For an overview of the structuring approaches, please see 4.2
Select the structuring approaches. This should, in theory, enable the user to identify the
17 In this case the NSC we are testing is Kompetenzklassifikation (the skills classification used by the German PES).
1
2 3 4
DLV011 – Summary report on ESCO skills pillar pilot
17 of 28
skill s/he is looking for in a more accurate manner compared to browsing through the
full-text results [1].
Furthermore, the right panel of the application “Contextualisation/Collections” [4]
supports the search by providing additional skills the user might be interested in. These
were not identified directly using search keywords [1].
2. Skills picker screen:
Figure 6: Skills suggestion screen - select an occupation
1
6 2
DLV011 – Summary report on ESCO skills pillar pilot
18 of 28
Figure 7: Skills suggestion screen - evaluate skills
The skills picker screen enables users to receive suggestions of skills which might be
relevant for their profiles based on occupations they select. The user first selects one or
more occupations [1] and the structuring approach they want to test [2]. Based on this
input, the application suggests various skills to the user. The user may accept or reject [3]
each one of the skills depending on whether they match her/his profile.
As the user accepts/rejects skills, the algorithm adjusts further suggested skills based on
the selected structuring approaches [2]. The skills which the user already accepted are
listed in the left-hand field “Accepted skills” [4] and further suggestions alongside their
evolving score are listed in the right-hand field “Top suggestions” [5].
The user may also switch the suggestion mode to “Broader concepts” [6]. This allows
her/him receive suggestions from a list of concepts stemming from NACE or the National
Skills Classification (in this case Kompetenzklassifikation) instead of those stemming
from the ESCO skills pillar.
4.4.2. Testing approach
Each screen of the application using the following methodologies:
Testing of “search screen”: The ESCO team searched for skills coming from a job
vacancy and a CV (randomly chosen) using the search and filtering mechanism of
the testing application. Then they filled in a short questionnaire on the usefulness
of each one of the different structuring processes.
Testing of “skills picker screen”: The ESCO team extracted the skills from a job
vacancy and a CV (randomly chosen). Using the skills picker application, the
team accepted those skills which were part of the identified set and rejected the
3
4 5
2
DLV011 – Summary report on ESCO skills pillar pilot
19 of 28
others. The team evaluated the quality of the suggestions based on (1) the number
of skills assessed, (2) the number of skills identified and (3) the number of
rejected skills.
5. Results and lessons learnt
5.1. Skills hierarchy
The results of structuring skills with the hierarchy of Kompetenzklassifikation and
DISCO are similar since both classifications i) follow a similar approach to structuring
skill concepts, ii) use a hierarchical grouping of concepts and iii) use sectors of economic
activity to structure sector-specific skills.
Neither contains detailed descriptions of their skill groups that could clarify their scope,
which makes it difficult to assign ESCO skills to the respective groups. Furthermore, both
classifications use economic sectors to classify domain-specific skills. Due to this, they
cannot be easily used as a mono-hierarchical structure, where each skill fits exactly in one
skill group. Certain ESCO skills could be assigned to more than one skill group as they
are not mutually exclusive.
While DISCO provides a breakdown of non-domain specific skills,
Kompetenzklassifikation has only one group. DISCO is available in 15 EU languages,
Kompetenzklassifikation only in German.
The effort required to maintain skills hierarchy based on existing classifications (DISCO
and Kompetenzklassifikation) is low since their structuring approach is built specifically
for the context of skills.
5.2. Other domains as the structure
Another approach to structuring skills is by expressing how they relate to other types of
concepts. This can be particularly useful for filtered search. The existing relationship
between skills and occupations in ESCO is of this type: it expresses which skills are
normally relevant in the context of an occupation and thus allows the user to filter the
skills pillar by occupation.
Following desk research, the ESCO team selected and tested four domains in the pilot
project: Economic activities (NACE), fields of education and training (ISCED-F),
product domain (CPA), Work context (Kompetenzklassifikation).
The effort required to maintain skill structure based on other domains listed here is
average since their structuring approaches are not built specifically for the context of
skills. On the other hand, most of the classifications (NACE, ISCED-F, CPA) contain
exhaustive descriptions of concepts which are very helpful when identifying the correct
concept for each relation.
5.2.1. NACE
The ESCO team assigned the ESCO skills to the relevant NACE groups of economic
activities. The relation of ESCO skills to NACE group mean that tasks enabled by the
skill are performed in scope the specific sectors of economic activity. This would allow
users to filter on skills that typically contribute to a specific economic activity. “Load
animals for transportation” could, for example, be a skill relevant among others in the
economic activities “freight transport by road” and “animal production”.
DLV011 – Summary report on ESCO skills pillar pilot
20 of 28
As NACE is a hierarchical four-level structure, skills can be assigned to economic sectors
at different granularity levels. In the pilot, several transversal skills were assigned to all
NACE sections since they apply to all sectors of economic activity. There was hardly any
ESCO skill that could be related to only one single NACE section.
The definitions of the NACE sections and the indications of the scope of the section are
very useful to delimit the scope and be accurate in assigning skills.
5.2.2. ISCED-F
The ESCO team assigned skills to the domains of education and training. The relation of
OSCO skill to specific ISCO-F code indicates where these skills are usually taught.
“Load animals for transportation” could, for example, be a skill taught in education and
training programs in the field of “transport services” or “crop and livestock production”.
At the lowest level, ISCED-F 2013 contains 80 detailed fields. The results showed that
the ISCED-F classification is valuable and would be particularly useful when annotating
learning outcome descriptions.
5.2.3. Statistical Classification of Products by Activity
The Statistical Classification of Products by Activity (CPA)18 lists services and products
in 3,218 subcategories. The relation of ESCO skills to CPA group mean that tasks
enabled by the skill produce, provide or utilise given product of CPA. It proved difficult
to decide on the appropriate relation to skills. It could, for example, express that
knowledge of a certain product or service category is useful for developing a skill. For
example, knowledge of “road transport services of live animals” or of “dairy cattle, live”
is useful for developing the skill “load animals for transportation”.
5.2.4. Work Context
The Kompetenzklassifikation work context classification is limited to 156 concepts
which is a small number to assign to 13,485 ESCO skills, whereas it is usually not
possible to assign a skill to a single work context. The relation of ESCO skills to work
context means that tasks enabled by the skill are performed in the location indicated by
the work context. Skills are usually performed in a range of different locations. For
example, the skill “load animals for transportation” can among others be performed at a
"harbour/wharf", an "animal hospital" or at an "animal enclosure".
5.3. Contextualisation
Skill contextualisation is a method to create knowledge or skill and competence concepts
by analysing how transversal skills, competences or knowledge are applied in the specific
context of a sector or an occupation. This allows bringing transversal knowledge, skills
and competences which are rather abstract to a more detailed level so that they can be
directly used in occupational profiles.
Contextualisation can be implemented through different relations that are part of the
ESCO data model. These are:
Broader/narrower relation (=hierarchical relation)
18 http://ec.europa.eu/eurostat/statistics-
explained/index.php/Glossary:Statistical_classification_of_products_by_activity_(CPA)
DLV011 – Summary report on ESCO skills pillar pilot
21 of 28
“Essential for” relations (=associative relation)
“Optional for” relations (=associative relation)
These types of relations can be established if the description of the skills is clear enough
to define its content and scope. The approach was used when developing ESCO v1, but it
has not been applied consistently over the entire skills pillar. Therefore, the ESCO team
proposes a set of rules for improving the skills descriptions. This will allow creating more
precise and consistent relations.
The effort required to maintain contextualisation structure is high since it requires a
detailed evaluation of each ESCO skill and (potentially) their relation to all other skills in
the classification.
5.3.1. Current status
Currently, 9,618 skills (71.3%) have at least one relation to another skill. In the figure
below, you can see a breakdown of skills per number of their relations.
Figure 8: Breakdown of ESCO skills with contextualisation
36
72
2
23
5
13
76
9
34
5
18
2
85
2
05
1
13
6
2
63
3
6
21
1
5
7 9
5
5
6
4
1
5
3
2
3
1
1
1
1
3
2
1
1
1
1
2
1
1
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
10
100
1000
10000
1
3
5
7
9
11
13
15
17
19
21
23
26
29
32
34
38
43
49
58
61
76
11
6
12
6
19
8
27
7
NU
MB
ER O
F K
SC
NUMBER OF RELATIONS
NUMBER OF KSC WHICH HAVE CERTAIN # OF RELATIONS
DLV011 – Summary report on ESCO skills pillar pilot
22 of 28
5.3.2. Proposed rules for contextualisation
5.3.2.1. Definition of terms
Skill features
• Skill features represent distinct parts that a skill should contain. These are:
• Action [A]: an action performed within the scope of a concept
• Subject [S]: the subject of the action which is performed
• Context [C]: the context in which the action is performed
Example:
work in metal manufacture teams – A: Work; S: In teams; C: metal manufacture
evaluate information – A: Evaluate; S: Information
plan weapon use on stage – A: plan; S: weapon use; C: on stage
advise on poisoning incidents – A: advise on; S: incidents; C: poisoning
Skill A is Broader (BT) to Skill B – Skill B is Narrower (NT) to Skill A
• Skill B represents a specialisation and/or a part of Skill A
• Skill A is more general in all of its features (Action/Subject/Context) to Skill B
In the following examples, we mark with red colour those relations which refer to an
incorrect case, something that should be prevented in the classification. We mark with
green colour the correct ones.
Example: manage data -- BT --> gather data
Explanation: “gathering” is a part of “management”, therefore the relationship is
established correctly
Skill A is Essential (ESS) for Skill B:
• Skill A is required in order to carry out Skill B
• Not possible to establish if a hierarchical relation already exists, e.g. "Skill A is
Broader to a Skill B".
Skill A is Optional (OPT) for Skill B:
• Depending on the context, Skill A may be required or be beneficial in order to
carry out Skill B
• Not possible to establish if a hierarchical relation already exists, e.g. "Skill A is
Broader to a Skill B".
DLV011 – Summary report on ESCO skills pillar pilot
23 of 28
5.3.2.2. Proposed general rules
The concepts have to be analysed holistically, the analysis should not be solely based on
the preferred term of the concept.
Example:
collection management -/- BT -/-> collection management software
Explanation: The subject of the first skill is a process and of the second a software, since
the subject is different the relationship cannot be established
manage data -/- BT -/-> manage data collection systems
Explanation: The subject of the first skill is a data/database, and a method/strategy for
the second one since the subject is different the relation cannot be established
In case a hierarchical relationship already exists, it takes priority over the associative
relation.
Example:
write work-related reports -- BT --> write stress analysis reports
write work-related reports -/- ESS for -/-> write stress analysis reports
Explanation: The first line establishes a hierarchical relationship, which has priority
over the second line, rendering the second incorrect
5.3.2.3. Proposed rules for the hierarchy
In order to establish a hierarchical relation between two skills, each action, subject and
context of the narrower skill must be the same or narrower than that of the broader (or be
absent).
Broader skill
Narrower skill
Action
Action
Subject
Subject
Context
Context
Each feature should be evaluated and only if all of them follow the aforementioned rule, a
BT/NT relation can be established.
Example: create production schedules -/- BT -/-> adjust the production schedule
Explanation: The action of the first skill is to set up the initial schedule, while the action
of the other one is to adjust it. Therefore, a BT/NT relationship cannot be established.
DLV011 – Summary report on ESCO skills pillar pilot
24 of 28
In case a feature is missing in a skill, this skill acts as “the broader”.
Example: order products -- BT --> place orders for orthopaedic products
Explanation: The context of the broader skill is missing, therefore the concept is
correctly considered broader
Hierarchical relations can be established only from skills with higher reusability level19 to
lower or the same not vice versa:
1. Transversal (highest)
2. Cross-sectoral
3. Sector-specific
4. Occupation-specific (lowest)
Example: manage data -/- BT -/-> manage quantitative data
Explanation: “manage data” is a sector-specific skill, while “manage quantitative data”
is a transversal skill. A transversal skill cannot be narrower to a sector-specific skill,
therefore this relation is incorrect
A skill20 and a knowledge21 cannot be related hierarchically.
Example: orthopaedics -/- BT -/-> sell orthopaedic goods
Explanation: orthopaedics is knowledge concept, therefore it cannot be established as
broader
5.3.2.4. Proposed rules for associative relations
The relation ESS for can be established only if the context of the two skills is identical.
Example: animal welfare -- ESS for --> promote animal welfare
Explanation: “animal” context is identical between both skills, therefore the relationship
is correct
19 https://ec.europa.eu/esco/portal/escopedia/Skill_reusability_level
20 https://ec.europa.eu/esco/portal/escopedia/Skill
21 https://ec.europa.eu/esco/portal/escopedia/Knowledge
DLV011 – Summary report on ESCO skills pillar pilot
25 of 28
The relation OPT for should be established if all other relations are ruled out.
Example: food homogenisation -- OPT for --> interpret data in food manufacturing
Explanation: no other relationship could be established between these two concepts,
therefore OPT relation is correct.
Associative relations can be established only from skills with higher reusability level to
lower or the same not vice versa.
Example: components of clocks -/- ESS for -/-> maintain clocks
Explanation: “maintain clocks” is cross-sectoral skill while “components of clocks” is a
sector-specific skill. A sector-specific skill cannot be essential for a cross-sectoral skill.
A skill and a knowledge cannot be related associatively.
Example: speak different languages -/- ESS for -/-> foreign languages for international
careers
Explanation: “speak different languages” is knowledge concept, therefore it cannot be
established as ESS for skill concept
5.3.3. Identification of potential skills for contextualization
In order to streamline the identification of potential hierarchical and associative relations,
the ESCO team could develop a system which would suggest relevant skills based on:
• their co-occurrence in occupations;
• text similarity in all languages;
• co-occurrence in groups of external hierarchies and
• data from EURES users on the skills they use in job vacancies and CVs (future).
Such a system would be based on an algorithm supporting the mapping platform22 or
skills picker and could be installed as a module of ESCO taxonomy management service
(or another ESCO maintenance service). It is recommended to use such tool once the
structure of the skills pillar is improved.
22 Mapping platform is a central European IT platform that allows Member States to create, update, manage
different versions of mappings and publish correspondence tables (mappings) - free of charge. In short, the
platform serves as an enabling infrastructure to map other classifications to ESCO. It allows to manually
map the concepts from one classification to concepts of another classification. It is supported by advanced
search options, automatic mapping suggestions and user-friendly interface.
DLV011 – Summary report on ESCO skills pillar pilot
26 of 28
6. Recommendations
Following the results of the pilot, the ESCO team recommends further exploring the use
of ISCED-F and NACE as external classifications to structure ESCO.
Contextualisation is a useful approach since it provides clear relations among skill
concepts. It can be used for various purposes and the skills pillar is already annotated
with such relations. Furthermore, revising the contextualisation and the methodological
rules related to it would improve the consistency of the skill pillar. On the other hand, it
requires a high effort to manually maintain such relations due to a high number of
concepts in the skills pillar.
DISCO and Kompetenzklassifikation, the two hierarchical skills classifications
evaluated within the scope of this project, use sectors of economic activity to structure
skills. Sectors of economic activity are not suitable as a base for a hierarchical
arrangement of concepts in which each concept can have only one broader concept
(mono-hierarchy), which is the hierarchical arrangement of interest in the case of ESCO,
because a single skill can inherently be related to several different sectors of economic
activity.
6.1. Possible further testing
The ESCO team identified further approaches for testing to improve the structure of the
skills pillar. These are the following:
Machine-learning structure: develop a structure based on the co-occurrence of
skills in various documents on the web. This could be achieved once the
Commission services receive data input from partners who use ESCO, e.g.
EURES, Cedefop, etc.
Collections based on common words: common words could be used to group
skills, e.g. a collection named “management” could be composed of all skills that
contain the word “manage” in their terms or description.
DLV011 – Summary report on ESCO skills pillar pilot
27 of 28
6.2. Implementation into full ESCO
Implementation of each structuring approaches will require various effort indicated, as
indicated in each of the detailed sections (c.f. Results and lessons learnt). The process of
structuring can be supported by software tools for semantic alignments, such as the
ESCO mapping platform.
6.3. Maintenance in full ESCO
Maintenance may consist of several different aspects:
Update during the release of new version of ESCO
Update during the release of new version of an external classification used for
structuring
Update for correction purposes
6.3.1. Update during the release of new version of ESCO
Such update is inherent in the maintenance cycle of ESCO and has to be performed
accordingly. On the other hand, it can be expected that not all skills will require a new
structure assignment, it is required to only update skills which were impacted (added or
modified) in the scope of the new release.
6.3.2. Update during the release of new version of an external classification
In case an external classification used for structuring (e.g. NACE) is updated the ESCO
relations to it should be updated as well. The urgency of such update will be directed by
the adoption rate of the new version of the external classification. The update can also be
supported using conversion tables provided by the maintainers of the external
classification. Such conversion tables commonly list the concepts which were impacted
and how they relate to concepts in the former version.
6.3.3. Update for correction purposes
Identifying skills or attempting to correct problems in the full dataset of 13.485 skills can
be time-consuming and inefficient. Concentrating search and correction efforts to subsets
of the full skills list can make maintenance more efficient.
For example: following technological developments in the logistics domain, a Member
State identifies that some ESCO skills are missing their relation to NACE 5210:
Warehousing and storage. The Member State provides such feedback to the EC who
proceeds with a review of structuring in the logistics domain.
7. Open issues
The main pending issue identified during the pilot is the lack of an existing suitable
mono-hierarchy that could be used for statistical labour market data purposes. This could
be the default browsing mechanism of the skills pillar. Such classification would have to
be composed of skill groups which are mutually exclusive, i.e. a single skill can be placed
only in a single group, thanks to clear scoping of groups. For the ESCO occupational
pillar, ISCO serves as such a mono-hierarchical structure.
DLV011 – Summary report on ESCO skills pillar pilot
28 of 28
In case a mono-hierarchy of skills would be the most suitable structure to address the user
needs which are reflected in the use case examples, it would be necessary to build a
custom solution for the purposes of ESCO.