Advances in machine learning and natural language processing have enabled the development of advanced methods for processing textual data, including automated document anonymization. In public administration, documents that contain personal data are frequently shared or published, making the protection of sensitive information essential. Anonymization allows the removal or transformation of sensitive data while preserving the usability and structure of the document.
This thesis addresses the problem of anonymizing named entities in official documents using machine learning techniques. The aim of the research was to develop and evaluate a model capable of recognizing named entities in Slovenian texts and automatically anonymizing them. The research was based on the named entity recognition (NER) approach implemented using the spaCy library and adapted for the Slovenian language. The model was trained on an annotated corpus containing examples of personal names, organizations, and locations.
In the practical part of the thesis, a prototype application for document anonymization was developed. The application allows users to upload text documents, automatically detect sensitive entities, and replace them with anonymized labels. The performance of the model was evaluated using standard machine learning evaluation metrics, including precision, recall, and F1-score.
The results indicate that machine learning approaches can effectively support the anonymization of textual documents while preserving their readability and structural integrity. The developed solution provides a foundation for further development of automated anonymization tools that could be applied in public administration and other environments where personal data protection is essential.
|
