The WollyBi Project
The WollyBi project was presented as an application case of how Artificial Intelligence in general, and machine learning especially, was used to monitor in real time the trend of web vacancies. The proposed methodology and the model used was published in the important magazine Future Generation Computer Systems, published by Elsevier, wrote by the professors Roberto Boselli, Mirko Cesarini, Fabio Mercorio e Mario Mezzanzanica.
The authors of the article – researchers and professors of the University of Milan Biccocca and affiliated to CRISP – presented the approach developed to automatically classify the multilingual job vacancies present online, through the development of a model of machine learning. The methodology identified and the results obtained gave rise to different application contributes, both in terms of analysis and instruments, including the Italian WollyBi project and the European project for the Cedefop organization.
The importance of analyzing job ads on the web
The automatic analysis of job vacancies is today of crucial importance for having a detailed picture (from nations to individual municipalities), in real time and standardized, of the labour market trends in terms of occupations and skills. Through the classification on the ESCO European multilingual taxonomy it is possible to observe the labour market, identifying the skills most requested by companies for each profession (more than 300 professions observed), also identifying “emerging” professions and skills, navigating a historical series in Italy from 2013 to now, with more than 2.5 million unique vacancies identified.
The procedure
Once the machine learning model was developed, the research was divided into four points:
- The most important job search websites have been identified for each European country involved in the project, developing a ranking model to guarantee the reliability and quality of the analysed source;
- Thanks to the agreements with the sources, each website has been periodically analysed to identify and classify new announcements published, eliminating the duplicates;
- The machine learning algorithm has classified the vacancies on the standard taxonomy, while an Information Extraction procedure has identified the skills from the texts and linked them with the ESCO European classification standard;
- The information obtained is made available in a web portal for browsing the data on different dimensions, such as Territory, Sector, Skills and Professions, based on the needs of the end user (from the agencies to the individual citizen)
The applications of the project
The WollyBi and Cedefop projects were the main outputs of the methodology developed for Labour Market Intelligence. WollyBi is the first Italian observatory that classifies the web job offers. The WollyBi experience gave life to phase 1 of the Cedefop project, which instead stems from the specific request of the European agency Cedefop (European Center for the Development of Vocational Training) to develop, in a first phase, a prototype based on the model of machine learning at 5 EU countries – UK, Ireland, Czech Republic, Italy and Germany – to extrapolate data that could be used to support decisions, with the goal of supporting the definition of new policies for the European labour market. Phase 1 of the project was successfully completed in 2016. The subsequent phase 2 was launched at the end of 2017 and is currently underway. The CRISP researchers today lead an international team, in which experts from all 28 EU countries are involved, to create the first real-time European labour market observatory. The project will allow to observe the trend of the Web labour market in real time in all 24 official languages ​​of the European community.
[/fusion_text][/fusion_builder_column][/fusion_builder_row][/fusion_builder_container]