Machine translation has become critical for overcoming language differences in the age of global communication, especially in specialized fields. This article discusses the complex process of evaluating the adequacy of machine translation for written specialized literature. The study analyzes the mistakes made by Google Translate (GT) and DeepL Translate (DT) when translating biology textbooks. The paper emphasizes that specialized texts require not only terminological accuracy but also context preservation, which makes machine translation in this field extremely challenging. Several evaluation approaches have been considered in previous work, emphasizing the need for human editing alongside machine translation systems. Machine translation errors, ranging from terminological and semantic mistakes to stylistic and grammatical errors, emphasize the importance of thorough post-editing. The results of the study showed that despite the correct translation of commonly used terminology, its use in complex sentences caused difficulties. Both systems often failed to recognize details, leading to distorted interpretations. Stylistic mistakes, such as word-for-word translation disregarding context, further weakened the translated sentences. The translation was also accompanied by grammatical errors that demonstrate the complexity of syntactic structures in specialized topics. The study revealed minimal differences between GT and DT, with GT translating terminology better and DT demonstrating more efficient grammatical structures. However, both systems used word-for-word translation, which resulted in stylistic errors and inaccurate word choice. This paper emphasizes the need for human editing to ensure the correct translation of specialized materials, as well as the need for continuous improvement of machine translation technology to ensure the accuracy of terminology and preserve context in specialized texts.
Read full abstract