Back to search

Institutional Repository of Vilnius University

Netiesioginių asmens duomenų anonimizavimas lietuvių kalba, pritaikant didžiuosius kalbos modelius /

Abstract

dc:description

This Bachelor's thesis investigates the anonymization of indirect personal data in the Lithuanian language using Large Language Models (LLMs). The main goal was to adapt and evaluate a Lithuanian LLM for this task, considering language specifics and GDPR requirements. A synthetic dataset, anon-lthuania, was created, and a Llama2 model was fine-tuned using the LoRA method to produce llama2-anon-lt. This model was benchmarked against international LLMs (ChatGPT-4o, Gemini 2 Flash, DeepSeek-V3), and the impact of different prompt types (Zero-shot, Few-shot, Chain-of-Thought) was analyzed. Llama2-anon-lt proved competitive (achieving ~91% of the top competitor's Composite score), demonstrating high precision, while Chain-of-Thought prompts significantly improved anonymization quality across all models. The work confirms the potential of LLMs for the practical anonymization of Lithuanian indirect personal data.

Degree

thesis:*
Grantor dc:publisher
Institutional Repository of Vilnius University
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Norvaiša, Kristupas,

Rights

dc:rights
Statement dc:rights
  • info:eu-repo/semantics/openAccess
Language dc:language
lit

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
Vilnius University
Base URL
epublications.vu.lt/oai
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Norvaiša, Kristupas,. Netiesioginių asmens duomenų anonimizavimas lietuvių kalba, pritaikant didžiuosius kalbos modelius /. Institutional Repository of Vilnius University, 2025. https://repository.vu.lt/VU:ELABAETD239031954&prefLang=en_US