Skip to main content

I could not attend the launch webinar of ESCO v1.1, where can I find the recordings?

The recordings of the presentations done during the launch webinar of ESCO v1.1 can be found here.

Was the semantic similarity approach in v1.1 used with English only or did you use multiple languages?

The semantic similarity approach was built using English, but we are working towards a multilingual model. We are currently collecting multilingual data and we welcome the sharing of data in European languages.

Were job vacancies the only empirical source for updating the skill-occupation-linkages in v1.1.? 

Job vacancies were only one of the sources to build links between skills and occupations. The primary source employed is feedback from experts, such as sectoral associations, European projects, etc. Other sources include national classifications, non-EU taxonomies, and qualifications.

How did you identify obsolete skills in v1.1?

The methodology to identify obsolete skills aims at removing duplicate concepts from the ESCO taxonomy.

The ESCO team focused on the new transversal skills and compared them to existing ESCO skills using our semantic similarity models and human validation. More details can be found in the videos of ESCO  v1.1 launch event.

When will ESCO include skills levels?

ESCO aims at becoming a knowledge graph. The Commission is currently reflecting on the different metadata needed to accomplish this goal and will count on the ESCO governing body to provide guidance and support. Skills levels have been requested by many stakeholders and the Commission takes it as a high priority.

Are the transversal skills also part of the EURES regulation?

The skills pillar of ESCO is composed of skills, knowledge concepts, language concepts and transversal skills. According to the EURES Regulation, EURES countries have to map their skill classifications to the ESCO skills or adopt the entire ESCO skills pillar. 

How will ISCO translations be corrected and updated?

The ESCO team does not manage the translations of ISCO. This is done by Eurostat so the process is independent from ESCO. If you have feedback on ISCO translations, you can let us know at EMPL-ESCO-SECRETARIAT@ec.europa.eu and we will transfer your message.

How are you representing text (e.g. from occupations) as numeric vectors? Is it based on TF-IDF, averaged word embeddings or sentence embeddings from a pretrained model?

We use sentence encoding using a pre-trained model finetuned on occupational data (ESCO, job vacancies, etc.). More specifically, we use the language representation model called BERT (Bidirectional Encoder Representations from Transformers).

Are Member States obliged to map their national occupations to the new ESCO occupations in v.1.1? What is the foreseen timeline?

ESCO v1.1 will only be mandatory when the EURES coordination group gives a positive opinion. However, since ESCO v1.1 is necessary to keep pace with the evolution of the labour market, and also largely compatible with ESCO v1.0, it is recommended that EURES countries update the maps of their national classification(s) to the latest version of ESCO. This task will be facilitated with a delta of the changes (mostly additions of new concepts) between ESCO v1.0 and v1.1 provided by the Commission.

Is the new ESCO version 1.1 already in the ESCO mapping tool/platform for Member States?

EURES countries who are using the mapping platform since 10th February (release of ESCO v1.1), can keep their current mappings to ESCO v1.0 but they will also be able to map their national classification(s) to ESCO v1.1.