Techniques for Transversal Skill Classification and Relevant Keyword Extraction from Job Advertisements

The recognition of transversal skills from job ads is important for ensuring a proper match between potential candidates and the requirements formulated in job ad texts. We contribute to understanding and interpreting job ad phrasings in two significant ways: firstly, we propose neural network-based classification models for the recognition of the six fundamental transversal skills formulated within the European Skills, Competences, Qualifications, and Occupations (ESCO) platform; secondly, we develop a means of identifying meaningful terms relevant to each transversal skill class, using feature importance-scoring methods that highlight the relevance of the words for recognizing each transversal skill. The resulting pipeline allows for the identification of skills in job ad texts, as well as the highlighting of important key terms for each recognized skill, therefore contributing to a better understanding of the skill taxonomy as well as the correlation of the related skill base with the corresponding formulations from job ads.