A pseudonymization tool for Hungarian

. In today’s world, the volume of documents being generated is growing exponentially, making the protection of personal data an increasingly crucial task. Anonymization plays a vital role in various fields, but its implementation can be challenging. While advancements in natural language processing research have resulted in more accurate named entity recognition (NER) models, relying on an NER system to remove names from a text may compromise its fluency and coherence. In this paper, we introduce a novel approach to pseudonymization, specifically tailored for the Hungarian language, which addresses the challenges associated with maintaining text fluency and coherence. Our method employs a pipeline that integrates various NER models, morphological parsing, and generation modules. Instead of merely recognizing and removing named entities, as in conventional approaches, our pipeline utilizes a morphological generator to consistently replace names with alternative names throughout the document. This process ensures the preservation of both text coherence and anonymity. To assess the efficacy of our method, we conducted evaluations on multiple corpora, with results consistently indicating that our pipeline surpasses traditional approaches in performance. Our innovative approach paves the way for new pseudonymization possibilities across a diverse range of fields and applications.