Kurnaz, AhmetUnver, Akin2026-02-032026-02-032025979-8-3315-6656-2979-8-3315-6655-52165-0608https://doi.org/10.1109/SIU66497.2025.11112347https://hdl.handle.net/20.500.12428/3468933rd Conference on Signal Processing and Communications Applications-SIU-Annual -- JUN 25-28, 2025 -- Istanbul, TURKIYEThis paper presents the practical application of big language models in text mining and topic modeling processes using texts shared online by the Ministry of Foreign Affairs of the Republic of Turkiye. The research describes an iterative and hybrid algorithm in which domain experts and large language models work together. The findings indicate that well-defined themes play a critical role in model success. It is emphasized that in social sciences, topic modeling processes should be handled with a tailor-made approach that includes the contextual interpretations of the researcher in the data evaluation processes, rather than being evaluated only through numerical outputs. In conclusion, the proposed algorithm offers a comprehensive and flexible analysis opportunity in social scientific research by generating topic labels suitable for qualitative or quantitative analysis of texts.trinfo:eu-repo/semantics/closedAccesstopic modellingLLMsnatural language processingConference Object10.1109/SIU66497.2025.11112347N/AWOS:0015754625003172-s2.0-105015463350N/A