Hundreds of thousands of online job vacancies (OJVs) were collected over 8 consecutive months in 2020 (April-December), processed and automatically classified against such international classifications / taxonomies as ISCED 2011, NACE, NUTS / ISO and ESCO. They provide unique insights on skills and occupational features and dynamics of the Tunisian and Ukrainian labour markets. Some of the many possible angles of analysis are visualised in the two countries’ dashboards, providing unique insights based on their granularity and real time nature.

The data system uses ESCO as the reference for machine-classification of skills identified in the hundreds of thousands of OJVs. For the case of Tunisian OJV data, the ESCO French and English versions were used. For the case of Ukraine, an additional step was indispensable: translation of ESCO skills into Russian and Ukrainian (over 4,000 terms).

The particular advantage of OJVs as sources for labour market intelligence lies in the fact that they present employers’ actual skill needs for the purposes of their business or activity in a given period.

The machine processing and classification of employers’ own terms and descriptions of skills shows cases of OJVs skills without a direct ESCO correspondence. This is due to the fact that a managed classification such as ESCO requires a certain time to identify, reflect and validate the integration of real time changes in the labour market. To overcome this challenge, the data science team involved in the project applied machine-learning techniques (e.g., Word2Vec) to enhance ESCO skills, creating a correspondence between a new ‘non-ESCO’ skill with a close (approximate) ESCO-skill. This process and the machine-proposed correspondence is discussed and validated by (human) experts in the given sector, occupation and technology.

The ETF project continues in 2021, with the addition of a new country and the continuation of data ingestion, classification and analysis for Tunisia and Ukraine. Many new questions and queries can be analysed on the basis of this growing database of OJVs - “let the data speak” - which are very useful in the context of ESCO’s continuous updates.

For more information about the project, please contact Eduarda Castel-Branco (ETF) and Alessandro Vaccarino (Burning Glass Europe).