Job offers classifiers for ISCO and KZiS 2023

Repository contains models used in the paper Multilingual hierarchical classification of job advertisements for job vacancy statistics How to use: https://github.com/OJALAB/paper-job-ads-classifierModelsLinear models:linear-bottom-2024.ziplinear-top-2024.zipTransformer models based on Polish data (trained with HerBERT-base and HerBERT-large)transformer-bottom-base-2024.ziptransformer-bottom-large-2024.tar.gz (with .partaa, partb)transformer-top-base-2024.ziptransformer-top-large-2024.tar.gz (with .partaa, partb)Transformer models based on Polish data (trained with xlm-RoBERTa-base and xlm-RoBERTa-large)transformer-bottom-base-rob-2024.ziptransformer-bottom-large-rob-2024.tar.gz (with .partaa, partb)transformer-top-base-rob-2024.ziptransformer-top-large-rob-2024.tar.gz (with .partaa, partb)Transformer models based on multilingual data (Polish and English; trained on xlm-RoBERTa-base only)transformer-bottom-rob-2024-argos.zip -- trained on dataset translated with Argostransformer-bottom-rob-2024-google.zip -- trained on dataset translated with Google Translatetransformer-top-rob-2024-argos.zip -- trained on dataset translated with Argostransformer-top-rob-2024-google.zip -- trained on dataset translated with Google TranslateTransformer models based on multilingual data (24 languages; trained on xlm-RoBERTa-base only)transformer-multi-bottom-2024.ziptransformer-multi-top-2024.zipExcel file with information about the quality of classifications:models-accuracy.xlsx