top of page
Writer's pictureSophia Behar

January Reflections

At school, in my last few Chinese classes before Winter break, we were working on a project involving the eight major cuisines in China. We learnt about the significant impact that geographical location has on the type of food in each cuisine. My teacher explained this through the idiom: “靠山吃山,靠水吃水”. At first, my classmates and I were perplexed given the literal meaning of the phrase is “rely on the mountains to eat the mountains, rely on the water to eat the water”, which does not make much sense. It turns out that the phrase actually refers to making the best use of the resources locally available, whether from the mountain or the sea. Our analysis of this idiom sparked my curiosity about machine translation. I began wondering whether online translators face the same struggles that we humans sometimes do, and resort to word-for-word translations or are capable of recognising the accurate, figurative meaning of idioms in different languages.

 

Hence, over the break, I decided to read about three popular translation services: Google Translate, DeepL and Microsoft (Bing) Translator. Google Translate is especially renowned, because it supports over 100 languages and even has a conversation mode, which is particularly useful when travelling. DeepL is a much newer translator that was launched in 2017 and is available in fewer languages but claims that it is “over 3x more accurate” than other translators. Lastly, Microsoft Translator offers some useful tools that allow someone to speak their native language in an online conference, with others hearing and seeing, as captions, what is being said in their own languages. Overall, each translator has been designed differently, with its specific perks. However, I wanted to experiment with them myself to witness them first-hand.

 

To start, I tried to translate some straightforward, literal text from Mandarin to English. I pasted in the paragraph that Ernie, the Baidu chatbot, had responded to me with when I asked it about computational linguistics in my December Thoughts blog post. This did not pose any problems to the translators, as each provided me with accurate and similar translations, and the only minor differences were synonyms, such as “experts” in one as opposed to “specialists” in another.

 

Next, I decided to take the difficulty level up a notch, testing the translators on idioms and sayings.

 

I first decided to ask the three translators to find the English meaning of the idiom I had learnt in Chinese class: “靠山吃山,靠水吃水”. While Google Translate and Microsoft Translate both offered the literal translation, DeepL suggested “fig. make the best use of local resources”. It was not only able to note that the sentence was a piece of figurative language, but also knew its underlying meaning without any context. 

 

Then, I wanted to see whether the “direction” of the translation had an impact on the translators’ accuracy. Hence, I requested that all three translators translate “it’s raining cats and dogs” from English into Chinese. This time only Microsoft Translate came up with the literal translation. Both DeepL and Google Translate explained that the phrase means it is raining heavily, and, interestingly, Google Translate even used a Chinese idiom “倾盆大雨” to portray the same meaning. 

 

Lastly, I was curious how the translators would perform if I changed the language. I put a French saying: “poser un lapin à quelqu’un” into all three machines, awaiting an English output. Once again, Microsoft Translate could not seem to decipher the metaphorical meaning, simply responding with “asking a bunny to someone”. On the other hand, DeepL and Google Translate clearly highlighted that the idiom actually refers to standing someone up.

 

Overall, all translators are extremely effective when it comes to more basic, literal pieces of text. Yet, according to my experimentation, when bringing in more complex, metaphorical sentences, DeepL’s claim that it is the “world’s most accurate translator” may not be an exaggerated one.


Credit: Prateek Joshi (Medium)


Works Cited


Alsan, Merve. “The Best Machine Translation Software You Can Try in 2023.” Weglot, 21 Sept. 2023, www.weglot.com/blog/machine-translation-software. Accessed 2 Jan. 2024.


Colin. “Best Alternative for Google Translate? 4 Top Options Compared in 2023.” TranslatePress, 4 Oct. 2023, translatepress.com/best-alternative-google-translate/. Accessed 2 Jan. 2024.


DeepL. “Why DeepL.” DeepL, www.deepl.com/en/whydeepl. Accessed 2 Jan. 2024.


Langster. “4 Best French to English Online Translators.” Langster, langster.org/en/blog/4-best-french- to-english-online-translators/. Accessed 2 Jan. 2024.

bottom of page