Поєднання мов програмування та великих мовних моделей у перекладі: дослідження швидкої розробки для налаштованої великої мовної моделі з використанням фреймворку COSTAR

“AI will continue to get way more capable and will become ubiquitous as time goes on. People are using it to create amazing things. If we could see what each of us can do 10 or 20 years in the future, it would astonish us today.” (Altman, 2024)

This quote comes from OpenAI CEO Sam Altman and reflects his optimistic view on the future of Artificial Intelligence (AI). The increasing capabilities of this technology have also made the application of large language models a hot topic. Particularly in machine translation, models like ChatGPT have demonstrated exceptional performance in translation tasks, becoming essential tools for professional translators. This study aims to enhance the quality of machine translation, especially for specialized texts such as patent documents, through prompt optimization and terminology extraction using the COSTAR framework. Special attention is given to the comparison between a customized GPT model, adapted with suitable prompts following the CO-STAR framework and targeted terminology optimization, and the regular ChatGPT-4 model. The influence of terminology and specific prompts on specialized translations is also considered to increase the efficiency and precision of machine translation.