How to Use Machine Translation

How to use Machine Translation
Please Share:

There’s no doubt that machine translation has made incredible strides in recent years. But it’s important to remember that it’s still just a tool- and like any tool, the value you get depends on how you use it. Here are our top 5 tips on how to use machine translation for your business.

Use MT Wisely

Machine translation is a boon to global businesses in a variety of industries. But it’s simply not the right choice for every situation.  For example, marketing and advertising content often uses cultural cues, jokes and figurative language to get the message across. That hooks consumers, but also means it’s often a poor choice for machine translation.  According to IBM, “Marketing slogans often rely on the audience’s understanding of a particular culture. Avoid these kinds of slogans because machine translation engines are likely to have trouble translating them accurately.”

When it comes to jokes, MT is much more likely to produce humour of the unintentional variety, as seen in this hilarious list of MT translation fails.

So, when can you use machine translation by itself?  Raw, unedited machine translation is a great choice for translating large amounts of content at once, in contexts where accuracy is not paramount. For example, user-generated content like reviews or comments is a common use case. Ecommerce product descriptions are another place where raw MT is often used.  (However, as frequent Amazon and shoppers can attest, this can have imperfect outcomes, especially if the MT engine has not been customised for your business.) For non-customer facing content, where you just want to quickly get the “gist” of something, MT is often the most logical choice.

But the more complicated or creative the content, the more problematic it is to rely on MT.  Advertising and marketing, in particular, often requires a different approach called “transcreation” to ensure the emotional impact of the original messaging doesn’t get lost in translation.

Customise your MT Engine

By customising your MT engine, you improve accuracy and generate more consistent translations. There are a variety of customisation options available, depending on the MT solution you’re using. For example, you may be able to edit previously produced translations to better “teach”  the MT engine what you want. Or, you may be able to upload translation memories with your desired translations for specific words and phrases. Most LSPs offer customised MT options and can manage this part of the process for you.

Prepare your Content for MT

To get the most out of MT,  your content should cater to the algorithm’s strengths. This may require editing the original text, prior to translating it. For example, machine translation performs better when the source content uses a consistent language and style.  Also, keep sentences short and simple. Between 5-20 words per sentence is the best practice, according to IBM.  But don’t make the sentences too short- machine translation engines do need some context. They also perform best when translating complete sentences, not phrases.  Remove unnecessary words, and write in active voice instead of passive voice whenever possible.  If you have trouble spotting overly complicated sentences on your own, an online editing program like Hemingway might be useful.

Be specific about your meaning- vague pronouns like “it” can lead to confusion.

Also, avoid using figurative language, slang, idioms or excessive acronyms. Otherwise, the resulting translations could leave your readers saying WTH and ROFLing- even if that wasn’t your intention.  However, if specific acronyms are part of your brand voice or essential to your business, this can be addressed in your terminology database.  Avoid Latin abbreviations like “i.e.”, “e.g.” and “etc.”

Finally, do a final spell check and proofreading of the original content before submitting it to MT. Errors in the source language can multiply in translation.

Human Supervision is Required

Since the advent of neural machine translation (NMT), MT accuracy has greatly improved. But even the most sophisticated MT technologies lack the nuanced understanding of a fluent, trained human translator.  That may not matter as much when you just want the “gist” of a translated document. But when accuracy matters, when anything and everything from your brand reputation to legal compliance to the lives and health of your customers is at stake, then MT must be accompanied by a robust quality assurance process and post-editing from qualified linguists.

Partner with a Quality Language Services Provider

How much is riding on the quality of your translated content? If you do business internationally, the answer is often “more than you think.” Even in something as mundane as a candy wrapper, translation errors carry a real and serious liability risk. (See this article to learn how an accidental translation of “nuts’ to “walnuts”  in product packaging put one company at risk of liability.) And even if nothing bad happens to a customer as a result, improperly-labelled products can be seized at the border, resulting in major financial losses. There’s also a definite reputation cost if consumer-facing content is mistranslated.

Machine translation offers real benefits in terms of speed and cost savings. But to take advantage of these benefits without putting your business at risk, you need an expertly-designed workflow that includes human post-editing from vetted translators and meticulous quality control procedures.

At K International, our team of project managers, linguists, multilingual copywriters, designers and compliance experts are standing by to make sure your projects are completed in the most cost-effective way possible without sacrificing quality.  For more information, contact us today.