uk_ner_gender-balanced

Model description

uk_ner_gender is a fine-tuned Roberta Large Ukrainian model, by using the same training configuration set as uk_ner_web_trf_13class Named Entity Recognition model for Ukrainian language.

This model was specifically trained on the NER-UK 2.0 gender-balanced dataset, a custom dataset designed to address gender bias in Ukrainian NER systems. The dataset includes both original and gender-swapped sentences, ensuring a balanced representation of male and female entity mentions, particularly in professional and role-based contexts. By training on this enriched dataset, ner-uk_for_gender-balanced_dataset aims to improve entity recognition performance across gendered language and reduce bias commonly found in Ukrainian NER models trained on unbalanced corpora.

Additional details can be found in the Github repository.

Usage

!pip install https://huggingface.co/linndfors/uk_ner_gender-balanced/resolve/main/uk_ner_uk_gender_balanced-1.0.0-py3-none-any.whl

nlp = spacy.load("uk_ner_uk_gender_balanced")