Ключевое слово: «machine translation»

In this paper, we present the first neural machine translation system for the Yakut language aimed at preserving and developing the linguistic heritage of the Sakha people. As part of the research, we have developed a model that considers morphological features of the Yakut language. Preliminary results show that the system can reproduce translations while preserving the grammar of the Yakut language. The developed translation system shows promising potential for practical applications in various fields. Further research will be aimed at extending the language model and improving the translation quality through additional training data and refinement of the algorithms. This work represents an important step in the development of computational linguistics for the languages of the peoples of the North of Russia and opens new opportunities for the preservation of digital languages.

Machine Translation (MT) has evolved from rigid rule-based systems to statistical methods and finally to neural architectures. While Supervised Neural Machine Translation (NMT) achieves high performance, it depends heavily on large parallel corpora. This survey reviews the techni- cal evolution of MT state-of-the-art (SOTA) approaches and analyzes Unsupervised Machine Translation (UMT). We examine how modern models learn translation mappings using only monolingual data, de- coupling performance from the availability of bilingual datasets.