Yes, I want my consultation

Machine Translation Is a Bad Master

Machine translation helps professional translators work faster and better. But as with many other things in life, it’s a double-edged sword that can be detrimental when used carelessly. Read this post for ideas about how to make the most of MT without falling into the numerous pitfalls it may set up for translators.


The general challenge about MT is that using it requires a higher degree of concentration and commitment than translating from scratch. Top translators who have these qualities will be able to benefit from MT. The rest will be easily deceived by what MT produces, which often looks good, but on the inside, it’s nothing more but a nice-looking guess that an MT engine made at the original meaning. And translators will keep this wrong translation intact while they would have likely translated it correctly if they had done it themselves, from scratch.


Another huge challenge about MT, the one that scares me most, is the influence MT will have on style. Because the style MT produces isn’t perfect, to put it mildly, the more you rely on MT, the more likely it’ll start to have a negative impact on your style. How does an individual generally form a feeling of style? By reading well-written materials such as fiction or non-fiction books and other quality publications. What happens to his/her feeling of style if an individual mostly reads junior high school essays? Right, it gets worse. The same happens to a person who lives in a country other than the county of his/her native language—he/she starts to mix up the two languages, and the style of the native one gets worse over time. In a similar vein, a translator who uses MT output will eventually come to believe its style is quite good. Here is an example from an English to Russian translation:

English: Finally, she gathers courage, asks, and is very glad she did.

MT output: Наконец, она собирает мужество, спрашивает, и очень рад, что она сделала.

While the word combination in bold may seem deceivingly good, it’s definitely not the correct one. The common phrase is “набирается храбрости” or “набирается мужества.”

Another example is that MT usually keeps the word order of a source language while a target language requires a completely different structure. This is especially true with creative texts. Restructuring sentences properly is a problem for many translators to begin with, and with MT arrival, I think it’ll get worse.


Most MT engines don’t have any idea about the importance of consistency. Instead of using one translation for a term throughout a document, an engine might use three or five. As a result, for any translators who aren’t very meticulous in this respect, MT will make things worse. Without MT, there’s a higher chance they’ll use the same word they used before even when they don’t think about it, i.e. subconsciously. But when they rely on MT, their subconscious doesn’t have a say, and the chances are they’ll keep all the different translations MT is suggesting. Try a challenging word such as “performance” in different contexts to see what I mean.

Technical Errors

MT engines may make technical errors such as changing upper case to lower case or deleting something they consider irrelevant (sounds funny, but it does happen). These issues take more time to spot and fix than translating from scratch. And if a translator fails to notice this kind of an error, but a client does, this will be quite embarrassing.

Watch Out

These are some of the reasons MT can be a translator’s enemy rather than ally. Can you relate to my experiences with machine translation issues? Please share what you think in the comments.

And of course I can’t help wrapping up this post with one of those examples that make people believe MT is no good at all:

English: Click on Take me to my inbox to get to the mailbox.

MT output: Нажмите на Возьмите меня в моем почтовом ящике чтобы добраться до почтового ящика.

Back-translation: Click on Have me in my inbox to get to the mailbox.

This is the last post this year. We wish our readers a prosperous new year. Thank you for reading our stuff and commenting in 2012. This post about the future of translation industry generated the largest number of comments, by the way. Happy Holidays!

One comment

  • За единообразие на МП-выходе отвечает не система МП, а человек (оператор), “отвечающий” за создание и ведение словаря. Регулярная настройка словаря МП по заданной тематике позволяет обеспечивать высокий уровень единообразия – говорю так с учетом своего опыта (20 лет применяю МП на практике).
    Причем, я говорю о единообразии на уровне терминологии и элементов стиля.

    Да, термин “performance” – очень своеобразный в плане вариантов перевода. Но, на уровне словосочетаний очень часто имеет уже один вариант. В моем словаре для МП (англо-русский, Телеком + ИТ) – почти 200 устойчивых словосочетаний с этим термином. При таком уровне настройки словаря МП выдает очень высокий уровень единообразия терминологии (я бы сказал, в диапазоне 90-95%) еще до начала пост-редактирования. Почему не 100%? По ряду причин. Но, об этом расскажу в следующий раз. 🙂

    Спасибо за материалы!

Add comment

About the Author

Roman Mironov
Roman Mironov
CEO & Founder

As the founder of Velior, Roman has had the privilege of being able to turn his passion for languages into a business. He has over 15 years of experience in the translation industry. Roman has helped dozens of clients increase sales by making their products appealing for speakers of other languages.