Translation accuracy of computer terms by ChatGPT, Gemini, DeepL, and Google Translate from English to Indonesian

Abstract

This study evaluates how accurately four freely accessible translation systems (Google Translate, DeepL, ChatGPT, and Gemini) render 67 computer-specific terms from English to Indonesian. All 67 terms gathered from Word 2016 for Professionals for Dummies were translated by each system. Prompts were used to create simple sentences in which the specific terms were employed. The translations were then rated independently and quantitatively using Nababan’s translation quality assessment model theory which assesses translation accuracy into three categories, accurate, less accurate, and inaccurate.  Of the 21 data sampling analyzed, Google Translate produced 16 accurate translations, 4 less accurate translations, and 1 inaccurate translation. DeepL produced 15 accurate translations, 6 less accurate translations, and 0 inaccurate translations. ChatGPT produced 18 accurate translations, 3 less accurate translations, and 0 inaccurate translations. Gemini produced 18 accurate translations, 3 less accurate translations, and 0 inaccurate translations. Then, it can be concluded that both ChatGPT dan Gemini outperform Google Translate and DeepL by average accuracy score of 2.85 compared to 2.71.