VEUCTOR: training and selecting best vector space models from online job ads for European countries (2025)

The dataset contains word embedding models trained on online job advertisements collected from multiple European countries, developed within a research project aimed at the comparative analysis of skills and occupations across national labor markets.  The main objective of the study is to assess how different training configurations affect the ability of embedding models to coherently represent labor-market-related semantics, and to identify, for each country, the best- and worst-performing models according to both intrinsic and extrinsic evaluation metrics.  The shared data include the models selected as “best” and “worst”, their training hyperparameters, and the associated quality indicators.  The dataset is organized by country and supports reproducible analyses and methodological comparisons in the fields of skill intelligence, labor market analysis, and computational social science.