A language modelling approach for discovering novel labour market occupations from theweb

This article presents an approach for the identification of potential new occupations, i.e., professions, not yet codified by the international standard taxonomy ISCO. This work is framed within the research activities of theWoLMIS project, developed by the University of Milano-Bicocca for the CEDEFOP European Agency, which classifies on-line job offers according to the ISCO taxonomy by using machine learning techniques. The proposed approach is based on text analysis, in particular on the use of language models, and provides two main contributions in the labour market context. First, it can support labour market experts in identifying new potential occupations and the process of updating the ISCO taxonomy. Second, language models are an effective way to identify the most similar occupations to a given one (either new or already coded in the taxonomy) in terms of skills and competencies. The proposed approach has been tested on a dataset of English job vacancies, obtaining promising results.