Machine Translation Customization with a Focus on the English-Czech Language Pair

(English) Machine translation (MT) can now be considered a common part of specialized translation. Neural machine translation can be customized, i.e. adapted to the requirements of certain texts. The aim of this thesis was to determine whether customization affects the MT output in the case of three texts and, subsequently, how the tools differ in this respect. The thesis focuses specifically on terminology. In the light of current technological development, three tools were selected: the neural machine translator Deepl Pro, the large language model adapted for translation Phrase Next GenMT, and the large language model OpenAI o1. The terms were entered into the DeepL Pro and Next GenMT directly in the user interface. For the OpenAI o1, customization was performed using a prompt. The results, analyzed using complex statistical methods, showed that customizing the models significantly increases the probability of correct term translation. However, no statistically significant difference was found between the tools, and different texts also had no effect on the success of terminology translation. Klíčová slova (anglicky) translation, english, czech, machine translation, customization, artificial intelligence, AI, CAT tools