Nowadays, enabling the transmission of knowledge and information among people belonging to different cultures is fundamental. Therefore, there is an always rising request of translations. To address this rising demand, translators in the past decades have started to rely more on tools that can aid their job: machine translation (MT) and computer-assisted translation tools (CAT), which are now essential. However, when it comes to languages that are quite distant from each other, such as Russian and Italian, finding words in the SL that do not have an attested translation in the TL can create some difficulties. In this case, one of the strategies that can be adopted is transliteration. Even MT has to perform transliteration when processing those words, such as proper names and toponyms. One of the main problems is that the outputs across various MTs do not always correspond, producing TWs with different characters that do not allow for a standardized transliteration. In this thesis an analysis on the accuracy of MT outputs is provided, so to assess which perform better among Language Waver, ModernMT, Intento, DeepL and Yandex when transliterating Russian anthroponyms and toponyms into Italian, taken from Russian passports. The metric of the assessment is based on a comparison between the output of each word generated from MTs and the norms suggested by the Russian Ministry of Foreign Affairs for anthroponyms and the UNGEGN Working Group for toponyms. Moreover, further investigation is devoted to ChatGPT-4Omni and its processing of selected STs to assess the translation outputs produced after receiving specific instructions in the prompt.
A comparative study of machine translation engines: transliterating Russian anthroponyms and toponyms into Italian
GASTALDELLO, OLGA
2023/2024
Abstract
Nowadays, enabling the transmission of knowledge and information among people belonging to different cultures is fundamental. Therefore, there is an always rising request of translations. To address this rising demand, translators in the past decades have started to rely more on tools that can aid their job: machine translation (MT) and computer-assisted translation tools (CAT), which are now essential. However, when it comes to languages that are quite distant from each other, such as Russian and Italian, finding words in the SL that do not have an attested translation in the TL can create some difficulties. In this case, one of the strategies that can be adopted is transliteration. Even MT has to perform transliteration when processing those words, such as proper names and toponyms. One of the main problems is that the outputs across various MTs do not always correspond, producing TWs with different characters that do not allow for a standardized transliteration. In this thesis an analysis on the accuracy of MT outputs is provided, so to assess which perform better among Language Waver, ModernMT, Intento, DeepL and Yandex when transliterating Russian anthroponyms and toponyms into Italian, taken from Russian passports. The metric of the assessment is based on a comparison between the output of each word generated from MTs and the norms suggested by the Russian Ministry of Foreign Affairs for anthroponyms and the UNGEGN Working Group for toponyms. Moreover, further investigation is devoted to ChatGPT-4Omni and its processing of selected STs to assess the translation outputs produced after receiving specific instructions in the prompt.File | Dimensione | Formato | |
---|---|---|---|
GASTALDELLO_OLGA.pdf
accesso aperto
Dimensione
1.05 MB
Formato
Adobe PDF
|
1.05 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/78819