Machine Translation has improved a great deal in recent years.
Choose the right time and place to use it, and you’ll see significant business benefits.
You still can’t rely on Machine Translation software alone to solve your language challenges…
But when paired with the right experts, it can save you both time and money.
What is Machine Translation?
Machine Translation is software designed to translate spoken or written text from one language to another. It is sometimes abbreviated as MT or referred to as a Machine Translation engine.
There are many types of Machine Translation:
- Rule-Based Machine Translation – the original type of MT engines performs translations based on set rules.
- Statistical Machine Translation – by analysing large amounts of data, SMT engines use algorithms and statistical models to perform translations.
- Neural Machine Translation – NMT perform translations using deep learning.
- Hybrid approaches – mixes of the above.
What all these have in common are the pros and cons of Machine Translation which makes it unsuitable for certain circumstances. And absolutely perfect for others…
What are the advantages of Machine Translation?
Machines can do many things much faster than humans. Translation is one of them.
If time is a factor, or you have a huge quantity of material which needs to be translated, MT is definitely something you should consider.
A human translator might be able to translate 2000 words per day. A computer can do the same in minutes.
That’s not to say that a computer will be able to do so as fluently or accurately as a human. That’s one of the major disadvantages of MT which we’ll look at in a minute.
But the time reduction is a massive advantage when you need to translate certain types of content.
This is already being taken advantage of in the linked field of Computer-Assisted Translation. Translation Memories – which are essentially databases of already-translated terms – are a standard tool for most professional translators these days.
They serve to increase efficiency as well as to increase the cohesiveness of terms used within and across projects.
Google Translate might be free. But that’s not the kind of professional Machine Translation software that we’re talking about here.
Yet still, the cost of Machine Translation can be very low indeed.
Certain steps do need to be taken before an MT engine can be said to be truly cost-effective though. Otherwise, the quality of output isn’t going to be there:
- An engine needs to be trained on high-quality bilingual data – preferably relating to the field in question.
- That data needs to be cleaned, structured and processed.
- The engine also needs to be tested and have its output assessed by human post-editors.
Those tend to add to the initial cost of setting up an MT engine. The savings come later when huge quantities of content can be translated to a set standard in a matter of minutes.
3) Research applicability
One of the many uses which MT engines are put to is widespread data collection in multiple languages.
For instance, a news team might have an MT engine which translates news stories from around the world and searches them for relevant topics and developments.
The advantage of using machine translation for purposes like this is that all the engine is doing is showing you where to look.
You’ll already be aware of the possibility of translation error by an automated system working over such a broad range of topics. But, as the output isn’t going to be seen by anyone other than you, you won’t need any post-editing.
4) Useful for first drafts
Professional translators already use Translation Memories and other computer-based aids to assist them in the translation process.
But that’s not where Computer-Assisted Translation stops. Some translators – depending on the topic, language pair and type of content – can use an MT engine to create the first draft of a given project.
By doing so, they essentially place themselves into the role of post-editors who know that a large amount of work will be required. Yet, they may also be able to save a great deal of time correcting the errors the machine has made as opposed to doing all of the work personally.
Translation Memories are used by pretty much all human translators. By making all already-translated terms and key phrases instantly replicable, they ensure consistency and greatly increase efficiency.
Machine Translation software effectively multiplies this to its greatest extreme.
Unfortunately, there are some downsides to this…
Disadvantages of MT
The above might have you thinking that the days of human translators are numbered.
Alas (or luckily, depending on whether you’re a professional translator), there are several disadvantages of Machine Translation which you’ll need human input to overcome.
By having an expert train your MT engine on high-quality bilingual data beforehand you can reduce or partially mitigate some of these issues. But they’ll still be present even on the most advanced neural MT engine after it’s been trained to the highest possible level:
1) Accounting for context requires careful training
Context is perhaps one of the most difficult things to train a machine to recognise.
Make no mistake, a machine can be taught to take into account the importance of context. But it is something which does require extensive training on that high-quality bilingual data.
2) Trouble understanding ambiguity
There are times when a translator will need to fall back on their own experience to provide an accurate translation because what is written is ambiguous.
A machine can’t do that unless it has a set rule for every single possibility which could arise.
3) Inability to comprehend framing or message
One of the most unsuitable types of content to have machine translated is advertising copy, marketing slogans and the like.
That’s because a machine tends to be entirely literal in any translation. On the other hand, successful advertising depends – in large part – on creating some kind of emotional reaction in your audience.
A direct translation of any marketing catchphrase will rarely succeed. Most slogans don’t really make sense on a literal level. They evoke a feeling.
Here’s a good example. This is the slogan of Haribo, the popular confectionery, in various languages:
- English – “Kids and grown-ups love it so, the happy world of Haribo”
- German – “Haribo macht Kinder froh, und Erwachsene ebenso” (“Haribo makes children happy, and grown-ups too”)
- French – “Haribo, c’est beau la vie – pour les grands et les petits” (“Haribo, life is beautiful – for grown-ups and children”)
Even if you don’t speak French or German, you can see how the syllables fit might into the catchy jingle now they’ve been translated with a little license, in a process called transcreation.
That’s not something which a machine can do. At least, not yet.
There is a strong possibility that your translated copy may be over-literal, boring or confusing when you use MT in this fashion. Highly embarrassing – and costly – mistakes can also be made.
You only need to check out any given “101 translation mistakes” article to see what happens when things go really wrong.
4) No understanding of cultural references
Cultural references in any content should always be carefully localised. This is something else that a machine cannot do without help.
5) Machines are often over-literal
It’s no surprise that a machine will tend to translate in a very literal manner. There are some occasions where this might actually be an advantage.
But in most circumstances, you’ll want the understanding of implication that only a human translator or post-editor can assure you of.
How to use machine translation
Use with human involvement” is a warning which should be written on the tin of any MT engine.
Even Machine Translation software which has been properly trained on the right bilingual data should always have its output reviewed by a human post-editor if it is intended for any kind of external purpose.
But, thanks to recent advances in machine translation technology, there are an increasing number of situations – from internal company communications to the translation of reviews to international research – where MT can save you large amounts of both time and money.
Do you need to know more about whether machine translation might be right for your company or project?
Leave a comment below or contact us directly. We’ll be happy to offer you any advice you need.