Vietnamese to English translation projects: Potential traps and tips
Two months ago, we were assigned an interesting Vietnamese to English translation project by a well-known LSP. The end-client is a top brand that everyone on the Internet knows. The volume was more than 100,000 words. Certainly, this is not the first rst Vietnamese-English project we have done. However, due to its uniqueness, it could be a very good example of how Vietnamese source text may challenge your linguists.
Table of Contents
First, let’s take a look at the project background. Unlike other common large volumes projects referring to a single domain, in this Vietnamese to English translation project, the end-client has spidered the web and randomly pulled sentences from hundreds and thousands of websites. To be more specific, the content could be a segment from a newspaper, Twitter feeds, blogs, technical documents, hotel pages, user manuals, forums, Terms & Conditions and so on.
They have then provided us with all these random segments and want us to translate them. Consequently, no two-segment is in the same context. Every segment is a new sentence and it has no link to other segments. The quality requirements were very high, 2 LQA rounds would be performed during the project and translators must be English native speakers to ensure smooth and natural translated texts.
You may ask what the point is. That is a good question as we had to double-check as well. The whole task is part of a huge R&D (Research & Development) project funded by one of the world’s largest enterprises. The purpose is to feed analytical data into a machine to train it like how a human linguist would translate these random sentences.
Right after the project kick-on, we get our native English translators to work on around 300 rst segments as an initial sample work. And not wait until our client’s rst LQA, we carried out an internal assessment on those rst translations. Unfortunately, the quality was not up to our standard. We analyzed the main causes of the sub-standard quality and then listed them out. And they’re potential traps that you may encounter as well in any of your Vietnamese to English translation projects:
In Vietnamese, there’re many homographs, i.e. words pronounced the same but have different meanings and origins. Leaving those words alone, we, as native Vietnamese speakers, can’t even tell them apart. If there’s a short context, such as a sentence or a phrase, we can “sense” it with our native tongue. But normally it’s not the case for native English speakers.
Compound words, which make up a big portion of the Vietnamese language, are the second culprit. Compound words do exist in English as well. But in English, commonly two words or more are joined together without any extra symbol/space to create a new meaning (Rey, softball, redhead, keyboard, makeup, notebook), or joined together by a hyphen (daughter-in-law, over-the-counter, six-year-old). In Vietnamese, all compound words are in open form. It means that there is no hyphen in between and those compound words look perfectly like two or three single meaningful words. Consequently, understanding the translator on the source text could be dramatically affected and confusion could happen anytime.
Slang which is more typical for this R&D project should be listed anyway. For slang, of course, it requires translators to have a high level of exposure to informal contexts. If translators only have experience with technical domains, they fail without a doubt.
Sentences with no standard structure: this is a typical trait of the Vietnamese language in general. The Vietnamese language has a grammatical structure. However, a lot of sentences can be understood well without any standard structure. This happens mostly in spoken language or informal conversations. Again, the “native Vietnamese” factor determines if a sentence without standard structure could be understood correctly or not.
So we were in a dilemma: if the native English translators do the translation, then very likely they will not understand the source Vietnamese text correctly, which leads to severe mistranslation. But if we use native Vietnamese translators, we can not guarantee a smooth and natural translated English text. Equally important, it is against the project’s policy to use native English translators.
1. We still let a native English translator do the translation in the rst round.
2. Then a native Vietnamese reviewer will perform the review round. Please note it is “review”, not “revise”. Our Vietnamese reviewers will check if the translation is correct to the source Vietnamese text. If they detect any error or potential error, the English native translator will be noticed. Most of the time, the reviewer helps to explain how the source Vietnamese text should be best understood. We also check low-level errors such as punctuation or spelling during this round.
3. The translator implements changes where necessary. Overall, we don’t use Vietnamese reviewer to review the translated text. In fact, we use English linguists to finalize the text. All processes can be done easily on a bilingual review Wordle exported from Trados Studio. This simple solution worked beautifully! We passed both two LQA rounds by the client with 96/100 and 98/100 respectively.
Final thoughts on Vietnamese to English translation
Traps are always there while you work on a Vietnamese to English translation or any language pairs. However, if you deeply understand the language you’re working with, everything would become pretty much simpler. Would you agree?