Results of the summarizing process for the arXiv paper: 2206.10919v1

This study aims to replicate previous research that compared neural machine translations to human translations. The previous study found that neural machine translations contain more strongly-associated formulaic sequences made of high-frequency words, but fewer strongly-associated formulaic sequences made of rare words, compared to human translations. In this study, the researchers used a parliamentary corpus to see if the findings could be replicated. The text in the corpus was translated from French to English using three well-known neural machine translation systems: DeepL, Google Translate, and Microsoft Translator. The results confirmed the observations from the news corpus but with less pronounced differences. This suggests that using text genres that typically result in more literal translations, such as parliamentary corpora, may be preferable when comparing human and machine translations. Additionally, the study found that Google translations contained fewer highly collocational bigrams than DeepL and Microsoft translations. These findings provide insights into the differences between neural machine translation systems and highlight the importance of considering text genre when evaluating translation quality.

• Study aims to replicate previous research comparing neural machine translations to human translations
• Previous study found that neural machine translations have more formulaic sequences with high-frequency words, but fewer with rare words compared to human translations
• Researchers used a parliamentary corpus to replicate the findings
• Corpus was translated from French to English using DeepL, Google Translate, and Microsoft Translator
• Results confirmed previous observations but with less pronounced differences
• Suggests that using text genres resulting in more literal translations (e.g., parliamentary corpora) is preferable for comparing human and machine translations
• Google translations had fewer highly collocational bigrams than DeepL and Microsoft translations
• Findings provide insights into differences between neural machine translation systems and emphasize the importance of considering text genre when evaluating translation quality.

Researchers wanted to do a study that is similar to another study. The other study compared translations done by computers to translations done by people. They found that computer translations have more common words, but fewer uncommon words compared to human translations. The researchers used a big collection of documents from the government to do their study. They translated the documents from French to English using three different computer programs: DeepL, Google Translate, and Microsoft Translator. The results of the new study confirmed what they found in the previous study, but the differences were not as big. This means it's better to use certain types of texts when comparing computer and human translations. Google Translate had fewer pairs of words that often go together compared to DeepL and Microsoft Translator. These findings help us understand how computer translation systems are different and remind us that we need to think about the type of text when judging translation quality." Definitions- Neural machine translations: Translations done by computers using artificial intelligence. - Formulaic sequences: Groups of words that are commonly used together. - High-frequency words: Words that are used often. - Rare words: Words that are not used very often. - Corpus: A large collection of written or spoken texts. - Literal translations: Translations where each word is translated exactly as it is in the original language. - Collocational bigrams: Pairs of words that often appear together in a specific order.

Comparing Neural Machine Translations to Human Translations

In recent years, neural machine translation (NMT) systems have become increasingly popular for quickly and accurately translating text from one language to another. However, it is still unclear how these translations compare to those of humans. A recent study sought to replicate previous research that compared NMTs with human translations in order to gain a better understanding of the differences between them.

Previous Research Findings

The previous study found that NMTs contain more strongly-associated formulaic sequences made up of high-frequency words than human translations do, but fewer strongly-associated formulaic sequences made up of rare words. This suggests that while NMTs may be good at translating common phrases and expressions, they may struggle when it comes to more complex or rarer language.

This Study's Methodology

To see if these findings could be replicated, the researchers used a parliamentary corpus as their source material for this study. The text was translated from French into English using three well-known NMT systems: DeepL, Google Translate, and Microsoft Translator.

Results

The results confirmed the observations from the news corpus but with less pronounced differences. This suggests that using text genres that typically result in more literal translations—such as parliamentary corpora—may be preferable when comparing human and machine translations. Additionally, the study found that Google translations contained fewer highly collocational bigrams than DeepL and Microsoft translations. These findings provide insights into the differences between NMT systems and highlight the importance of considering text genre when evaluating translation quality.

Conclusion

Overall, this study provides valuable insight into how different types of texts are translated by both humans and machines alike. It also highlights the importance of considering text genre when evaluating translation quality as different genres can produce very different results depending on which system is used for translation purposes.