How convenient! But… is the quality of those machine-generated translations trustworthy?

Machine translation has constantly been improving but is still far from perfect. I would say that this is an era in which humans and machines have to work hand in hand to produce an adequate final product with an optimal price and TAT. In fact, this type of collaboration is becoming increasingly popular among LSPs under the name “machine translation post-editing,” or MTPE.

So, what is MTPE?

MTPE, or machine translation post-editing, is a process by which a source text is automatically translated by an engine, and the work is then revised by human translators. This process combines the unparalleled speed of machine translation with the deep linguistic/cultural knowledge of human editors.

Let’s have a look at an example of the MTPE process below.

It’s true that “blackouts” can be translated as “mất điện” (power outage) or “ngất xỉu” (faint). Still, in this medical context, the latter is clearly more suitable. Google Translate failed to notice this, and it’s the human editor’s job to fix this mistranslation.

Benefits of MTPE

One might argue, “Wouldn’t it be safer to rely on a human linguist for the translation step to ensure a usable target with minimal revisions needed?” While this traditional approach is still widely preferred by many language service providers (LSPs) for confidentiality and quality reasons, the MTPE process also has advantages.

• Saving time and effort: With the lightning speed of a machine, you can significantly cut down on the hard work put into the translation step. What once took hours of work can now be completed in minutes. This could mean that the translation cost will also be more affordable.
• Enhancing capacity: The average translation capacity of non-professional human linguists is about 2,000 words/day, while the average MTPE capacity ranges from 3,500 to 7,000 words/day. That is a massive boost in capacity.
• Improving translation quality: It’s safe to say that machines are more accurate than humans in the mechanical aspects of the translation process. Non-linguistic errors like typos, punctuation, space, capitalization, etc., are hardly seen on a machine-generated target. Furthermore, by having MT take care of the burden of translating tasks, human linguists can focus their energy on being creative and fulfilling clients’ requests.

Things to notice in MTPE projects

The biggest concern of MTPE projects is the target quality, which varies from engine to engine. Fortunately, we human translators have some tactics to control that quality.

• Choosing the right MT: MTs are not created equal. DeepL Translator does not support a wide range of languages, but it excels at those it does. Baidu Translate seems to be more friendly to Chinese-speaking users than that to other languages. Bing Translate can be ahead in literary translation in some cases. Depending on the language pair you are working with and the translation domain you are in, each MT has pros and cons. It’s best to test as many MT as possible to choose “the one”. Some companies even develop their own tailored, well-trained translation machine.
• Simplifying the source text: MT is nowhere near the language proficiency of human linguists, which means that it would not handle sentences that are too advanced in terms of syntax and lexicon well. To better understand how bad machines can be at translating tricky sentences, let’s analyze how Bing Translate transfers Demi Lovato’s famous lyrics into Vietnamese. Well, I can see that it’s too complicated for Bing to understand that the second “break” in the lyric does not mean “to shatter” (làm tan nát) but “to comfort” or “to heal” (xoa dịu). This example tells us that the source text used in MTPE projects should be simple in structure, evident in meaning, and error-free if we want an adequate translation.
• Paying extra care to the pattern of error of the MT: After years and years of being under improvement, MT still has not reached the expertise required to produce a perfect translation. Each translation engine has its own error pattern that post-editors should know. It may be a too-literal translation, terminology issues, or an unlocalized date format. The example below shows that Google Translate does not change the US date format (MM/DD/YYYY) to the Vietnamese date format (DD/MM/YYYY) if your DD is below 12 (if the DD is above 12, Google Translate does make the adjustment). Once you familiarize yourself with mistakes, your MT frequently makes, detecting and correcting them will be easier.
• Applying quality control measures: Built-in QA features in CAT tools can be beneficial when checking the accuracy of the MT. The memoQ QA feature alerts you if the terms used by the translation engine do not match the terms requested by your client. QA Checker on Trados displays warning messages when the machine mistakenly translates any brand names that are not supposed to be translated.

Why settle for slow, error-prone translations when you can have lightning-fast results with Hansem Vietnam’s MTPE services? Our team of expert post-editors works hand-in-hand with cutting-edge translation engines to bring you the best of both worlds – speed, and accuracy. With Hansem Vietnam, you won’t have to sacrifice quality for efficiency. So, why wait? Let’s translate and conquer!

References

• Andrea Gonçalves Pinto (October 22, 2019) 7 Things to Keep in Mind When Accepting/Working on a MTPE Job
  • https://www.linkedin.com/pulse/7-things-keep-mind-when-acceptingworking-mtpe-job-gon%C3%A7alves-pinto/?articleId=6592393715820576768
• 9 Best Practices for Machine Translation Post Editing (MTPE)
  • https://tarjama.com/9-best-practices-for-machine-translation-post-editing-mtpe/
• Muhamad Alif Haji Sismat (November 2020) Analysing Patterns of Errors in Neural and Statistical Machine Translation of Arabic and English
  • https://www.researchgate.net/publication/349160073_Analysing_Patterns_of_Errors_in_Neural_and_Statistical_Machine_Translation_of_Arabic_and_English
• June 17, 2021 What’s the difference between human translation and MTPE?
  • https://www.oneword.de/en/whats-the-difference-between-human-translation-and-mtpe/