DUTCH NAMED ENTITY RECOGNITION AND DEIDENTIFICATION METHODS FOR THE HUMAN RESOURCE DOMAIN

This paper evaluates Dutch named entity recognition (NER) and de-identification methods for the human resource (HR) domain, focusing on anonymizing privacy-sensitive texts. The study updates existing methods with advanced NER models, achieving high recall rates for suppressing personal identifiers, though it identifies areas needing improvement, particularly concerning gender and job title recognition. Additionally, a new dataset for recognizing job titles in texts is introduced, achieving a high F1-score of 0.91.