Machine learning translation
insights

Automatic Machine Translation versus Human Translations

Automatic Machine Translation has significantly improved in the past years. This begs the question: Is Machine Translation sufficient for a company website? Why would I still require and perform expensive manual editing?


Automatic Machine Translation ("MT") has significantly improved in the past years. This begs the question: Is Machine Translation sufficient for a company website? Why would I still require and perform expensive manual editing?

These are the questions that we ask ourselves at Dilingual every other day. It's our job to  enable everyone to translate their websites quickly and reliably. And Machine Translation is a huge factor in that.

How Machine Translation works

Artificial Intelligence or "AI" is a term that will inevitably come up when you read an article about Machine Translations. But what does AI mean? How does it work?

In a nutshell, modern neural machine translation is all about statistics. The so-called "model" is being trained by a special algorithm based on a corpus of input data.

The corpus of input data includes millions of translations from one language into another. The machine will learn the statistical likelihood of how a sentence is translated into the other language and apply these for all new sentences.

So there is no dark magic being used here; Machine Translation is good old-fashioned mathematics and statistics.

Advantages of Machine Translation

Machine Translation offers plenty of advantages, such as:

  • Instant translation of sentences into other languages
  • No resource capacities such as by human translators

Especially when you are dealing with marketing content, then being able to instantly have a translated version into other languages puts you ahead of your competition.

Limitations of Machine Translation

As Machine Translation is just a statistical model, it is far from perfect and will sometimes result in wrong or even misleading translations.

Some examples of issues that can happen with Machine Translation include:

  • Incorrect translation of formal/informal texts
    • In some languages, there is no informal/informal version. When texts are being translated into another language this can result in addressing the audience potentially incorrectly.
  • Incorrect translation of "false friends"
    • If the translation corpus is trained on data that includes these the translated texts could potentially also be mistranslated.
  • Incorrect translation of unknown words
    • If the text contains some unknown words then this results in the model having to guess. This can potentially result in wrongly translated words.
  • Lack of context over several sentences
    • Many Machine Translation solutions are translating only sentence by sentence. This results in a potential loss of context compared to the original text.

Verdict: Machine Translation versus Human Translation

Machine Translation has gone a long way in the last 15 years. It is however still not something you can rely 100% on.

That's why we recommend Dilingual users use our intuitive user interface to review machine translations of their website.

Doing so just takes a few seconds and will result in reliable and correct translations. In addition, any manual edit has the potential to improve future automatic translations for everyone.

Similar posts

Get notified about new blog posts

Subscribe to our newsletter to stay up to date on the latest translation insights and news.