Named Entity Recognition for Confidential and Military Texts in the Uzbek Language

This thesis presents a specialized Named Entity Recognition (NER) system for identifying sensitive entities in Uzbek confidential and military texts. By fine-tuning the bertbek-ner-uznews transformer model with domain-specific synthetic data, we achieved an F1-score of 89.6% across seven custom entity categories. The system addresses critical gaps in Uzbek NLP for security applications and provides a foundation for Data Loss Prevention (DLP) integration.