Slider

Making effective use of machine translation

 
Once upon a time, there was no such thing as Google Translate. Anyone that needed a translation had no other choice than to consult a translator or translation company.

Nowadays, machine translation technology has advanced to the point that many use it instead of hiring a translator. And while for some instances that works perfectly fine, for others it might not cut the mustard on its own.

So, how do you make effective use of a machine translation?
 

Know what to expect

The first thing to understand, and perhaps limit, before making use of machine translation are your expectations. So let’s have a quick assessment of what you’re likely to get.

Will it be quick?

Undoubtedly.

Will it be perfect?

Probably not.

But that is not necessarily a bad thing depending on the text’s purpose.
 

Based on our experiences of researching and working with machine translation, most systems tend to fall into the same following pitfalls. Some more frequently than others (you’ll find some fun examples at the end of this article).

  • Word choice for words with multiple meanings
    If your writing uses a word with several meanings, the context is very important for machine translation (as with human translators). However, while a human translator can use common sense to rule out some meanings, machine translation can only rely on algorithms and probabilities.
  • Noun genders
    We work with German and English, and one of the main differences between those languages is that in German all nouns come with a gender. So, even an inanimate object can end up being masculine or feminine. Unfortunately, this frequently poses a challenge for machine translation and can lead to some entertaining, if not problematic, errors.
  • Grammatical misinterpretations
    This type of error may often be linked to one caused by the gender system in German grammar. However, there are also some times when machine translation can trip over constructs such as separable verbs and other grammatical aspects.
  • Stylistic problems
    Understandably, machine translation struggles with more artistic aspects of writing such as rhythm, rhyme and metaphor. As such, if your source text has any of these incorporated into it, they won’t be in the machine-translated target text.
  • Oddities
    This last category of error is hard to define precisely and therefore can’t really be predicted or anticipated. It covers miscellaneous things such as random additional words.

 

Know when to use machine translation

As hinted at above, depending on a text’s purpose you may not need perfection. That’s why the second thing to understand about machine translation is when to use it.

An obvious use for machine translation is when you only need to know the gist of something. If you don’t ever intend to publish the document and just need to get a general understanding of what an email says or what a concept is, then machine translation is ideal.

Outside of giving you the gist, machine translation software is also now pretty good at simple, repetitive tasks. The ones that would often actually take a human considerably longer to do by themselves.

Take manuals for example. The source material is generally written in a very straightforward style and does not feature very much in the way of complex language. They also tend to need to be translated into many different languages. As such, they are a prime example of when machine translation can be a trump card to play. You’ll probably be able to rely on the translation relatively well too. Especially if you can have someone check them for any of the errors mentioned above.

Another way machine translation can be put to good use is as a first draft on documents to speed up the translation process. This gives the translator or proofreader a quick initial draft to work from, allowing them to tidy the text, applying any corporate and creative styles that may be relevant.
 

Hire a translator to proofread the document

The best way to ensure you get the most out of machine translations, as mentioned above, is to hire a translator to edit what the software gives you.

By doing this you’re guaranteed to capitalise on the best aspects of both methods of translating. You get the initial time savings that machine translation offers. But you also get a set of human eyes that can comb the text for any errors that may leave you with egg on your face.

What’s more, a translator is able to spot more than just surface-level grammatical errors in the target text.

Provided you give them access to the source text too, their bilingual nature will mean that they are well-positioned to correct any mistranslations. They will also be able to ensure that the document is a true reflection of the source material, uses accurate industry terminology and reads naturally.

Finally, a translator can adapt the text for your readership too. This could be by making sure any sensitive subject areas are dealt with properly and with care. Or that the text is culturally aligned. It might even involve reworking some aspects to use more accessible language, opening up your content to a wider audience.
 

Example mistakes

Here is a collection of just some of the funny results that machine translation created during our research* into using it, enjoy!

    Word choice for words with multiple meanings

  • Sleeping Beauty
    The crazy old woman…
    The princess enters the room and asks the old woman what she’s doing. She should reply that she is spinning. However, the German verb for to spin ‘spinnen’ can also mean ‘to be crazy’. So, machine translation handled the German section, “‘Was machst du da?’ – ‘Ich spinne,’ sagte die Alte und nickte mit dem Kopf.” as follows:
    ‘What are you doing there?’ – ‘I’m crazy,’ said the old woman, nodding her head.
  • Noun genders

  • Sleeping Beauty
    Sleeping Beauty magically turns into a man…
    Er brückte sich und gab ihm einen Kuss.
    He bent down and gave him a kiss.
  • Hansel and Gretel
    The forest becomes a man…
    ‘Weißt du was, Mann’, antwortete die Frau, ‘wir wollen morgen in aller Frühe die Kinder hinaus in den Wald führen, wo er am dicksten ist.
    ‘Do you know what, man,’ replied the woman, ‘we will lead the children out into the forest early in the morning, where he is the thickest.
  • Grammatical misinterpretations

  • Hansel and Gretel
    Setting the children on fire…
    Da machen wir ihnen ein Feuer an und geben jedem noch ein Stückchen Brot, dann gehen wir an unsere Arbeit und lassen sie allein. Sie finden den Weg nicht wieder nach Haus, und wir sind sie los.
    “…and we will set fire to them and give each one a piece of bread, then Let’s go to our work and leave them alone, they will not find their way back home, and we’ll get rid of them.
  • Snow White and the Seven Dwarves
    Organs as a symbol of the queen…
    Und als gerade ein junger Frischling dahergesprungen kam, stach er ihn ab, nahm Lunge und Leber heraus und brachte sie als Wahrzeichen der Königin mit.
    And just as a young young man was jumping along, he stabbed him, took out his lungs and liver and brought them with him as a symbol of the queen.
  • Stylistic problems

  • Three Little Pigs
    Not quite as poetic as “I’ll huff and I’ll puff”…
    Ich werde husten und prusten und dir dein Haus zusammen pusten!
    I’ll cough and flail and blow your house together!
  • The Boy Who Cried Wolf
    Goodbye rhythm and rhyme…
    Wer einmal lügt, dem glaubt man nicht, auch wenn er dann die Wahrheit spricht.
    Once you lie, you do not believe it, even if it then speaks the truth.
  • Oddities

  • The sleeping Sleeping Beauty…
    Es ging aber die Sage in dem Land von dem schönen, schlafenden Dornröschen
    But the legend in the land went of the beautiful, sleeping Sleeping Beauty
  • Snow White and the Seven Dwarves’ young young man…
    Und als gerade ein junger Frischling dahergesprungen kam, stach er ihn ab, nahm Lunge und Leber heraus und brachte sie als Wahrzeichen der Königin mit.
    And just as a young young man was jumping along, he stabbed him, took out his lungs and liver and brought them with him as a symbol of the queen.

 

In a nutshell

So to sum it all up, to get the most out of your machine translation:

  • Know what to expect – it will be fast, but probably not perfect
  • Know when and where to use machine translation
  • Hire a translator to edit the translation to achieve the highest level of quality

 
 
* Our original research was carried out in 2018. Given the rate at which these systems have been advancing, machine translation may now not make similar mistakes.