SEMANTIC AMBIGUITY OF URBANISTIC TERMINOLOGY (UKRAINIAN-ENGLISH GOOGLE TRANSLATE VS HUMAN TRANSLATION)
Abstract
The problem of semantics is one of the key concerns both for machine and human translation. Despite the fact that neural networks have been used as a central background for Google Translate, a human factor has contributed much to the final version of the translation. The optimal solution to the problems of semantics is the use of interactive mechanisms (dialogue, semiautomatic), which provide a mutual decision to the problem by a human being and the computer. The human factor of the translator is a decisive tool for quality translation achievement. Finally, the target text sounds as a source text, following all pragmatic intentions of the source text author.
This research considers the translation of an urban design article from Ukrainian into English (Google Translate vs human translation). Based on the analysis of 33,000 printed symbols, the semantic errors were identified and corrected. In addition to the semantic charges of words that Google Translate tries to correct using neural networks, the structure of the target language sentences does not meet the grammatical requirements of the English language in the analyzed passage. Machine translation seems to copy the structure of the original sentence and translates it not into English, but into "UkrEnglish". The translator based on his ‘sense of language’ restructures the sentence and in the process of translation he chooses the necessary lexical translation equivalents. Obviously, there are more bugs than Google Translate developers believe. Of course, the translation that appears is understandable in many cases but sounds unnatural in terms of the target language norms. Thus, in the passage examined, the translator made 244 corrections as can be seen from the Word automated comparison of the professional translation text and Google translate. Structural adjustments, semantic correspondences, a large number of pronouns, prepositions, and even articles have overloaded the Google Translate target text.
References
2. Ефремова Н. Нейронные сети : практическое применение. 2017. URL: https://habr.com/ru/post/322392/ doi:10.1038/nature.2016.20696
3. Іваницька Н. Повнозначне слово української мови в сучасних категорійних вимірах : монографія. Нілан-ЛТД, 2017. 266 с.
4. Castelvecchi D. Deep learning boosts Google Translate tool. 27 September 2016. URL: https://www.scientificamerican.com/article/deep-learning-boosts-google-translate-tool/
5. Lewis-Kraus G. The Great A. I. Awakening. Dec. 14, 2016. URL: https://www.nytimes.com/2016/12/14/magazine/the-great-ai-awakening.html
6. Melvin Johnson et al. Google’s Multilingual Neural Machine Translation System : Enabling Zero-Shot Translation. 2017. URL: https://www.aclweb.org/anthology/Q17-1024/
7. SCHUSTER M. ET AL. Zero-shot translation with GOOGLE’s multilingual neural machine translation system. November 22, 2016. Url: https://ai.googleblog.com/2016/11/zero-shot-translation-with-googles.html
8. Surafel M. Lakew, Quintino F. Lotito, Matteo N., Turchi, M., Federico, M. Improving Zero-Shot Translation of Low-Resource Languages. 2017. URL: https://arxiv.org/abs/1811.01389
ISSN 
.png)
