Jump to content

When Should You Use Machine Translation Services?

We may have pointed out the flaws of Google Translate this month, but that does not mean you should avoid machine translation.

We have pointed out the flaws of Google Translate in a previous blog, but that does not mean you should avoid machine translation. Search engine translation tools offer the quality you would expect for a service that's free. Although the tools are sophisticated and an impressive precursor to the future of translation, machine translation is not the same as a search engine tool.

Search engine translation tools use the internet in a risky fashion to create definitions. Machine translation is a much more sophisticated tool that notices regularities when comparing documents. The more you use machine translation, the quicker it builds up its bank of words and phrases and the more you save in the long run.

But translators shouldn't panic about any longer being in demand. Humans will always be needed to post-edit the work. Due to the translation being a creative medium, people continue to be its strongest asset. Machine translation quickens the process by building up a library of already translated material to allow the translator to focus on other sentences. In fact, the minimum improvement is a doubling in productivity – although it is more common to triple or quadruple.

Some questions do remain though. Such as, when should you use machine translation services? Also, is it right for your industry? And what do you think is best?

How Does it Work?

Technology is evolving at its fastest pace yet with many crazy things on the horizon (3D-printed cooked food, anyone?). Language is a little bit more difficult as languages aren’t entirely similar. Although there are more similarities between Mandarin and English than you might believe, human brains are wired to compute languages and indeed translate them through their brilliantly creative and puzzle-solving minds.

As technology picks up speed, though, it is starting to learn the intricacies of each language by building up banks and banks of data. The more computers are used, the more they will learn to understand – not unlike a toddler. MT is in its infancy, crawling at the moment before eventually (it's fair to presume) learning to walk, run, jump and dance the foxtrot. It looks for rhythms, statistics, and patterns to learn grammatical rules.

The assumption that machines will replace translators, though, is untrue as they will continue to work in a mutually beneficial partnership. Technology’s rapid development is not to be sniffed at and neither should the translators. As we say, both are an important part of the ever-changing translation landscape.

Coding to illustrate the intricacy of programming machine translation

Just Machine Translation

There are service levels, as with all translation, when it comes to machine translation. If you were to choose machine translation with no post-editing at all, you may ask: why not use a search engine tool? The problem with those is that they are not confidential. Your data is feeding into their database as it actively learns and could feed into your competitors’ pairings by learning yours.

By using the machine translation service of an agency, your database is individual for you and we sign strict-NDAs to ensure your data is secure and protected.

As it is also not edited, the translation will be extremely rough. This should only really be used for internal documentation if you need a quick way to understand something for, say, business meetings. Norbert, our man-in-charge when it comes to MT, recommends this for large volume work such as buyouts, takeovers, and similar situations.

The reason is, that if it’s for internal use, you only need the context of the content rather than the full deal with translation. As it isn’t facing outwards (i.e., your client/customer will never see it), you won’t damage your company’s reputation or brand in doing so, which allows you the freedom to look for this cost-and-time-effective solution.

Monolingual Post-Editing

This isn’t of the highest quality either, but it is a good middle ground. This is for global companies looking to communicate with branches across the globe in a clearer manner, but not in an exact fashion. This can be used in a variety of ways which depends on yourself. It's as simple as company announcements or even translation of training material.

Small instructional manuals are ideal for this form of MT due to their brief word count. If you have something under 500, the post-editor can turn around a quick correction of the piece to make sure that everything makes sense, ensuring your instructions are safe and coherent. Even simple things like clothes tags require translation and this form is the best as it only requires one post-editor. It's cost-effective, has a quick turnaround, and keeps your outward-facing material of high quality.

Bilingual Post-Editing

The difference is indeed in the title. The corrections are fed back into the machine so the machine then learns the error of its ways and banks them for the next time that sentence, phrase, or terminology is used. In time, this could lead to a 70-85% saving in the future, depending on how complex or consistent your terminology is in technical documents.

It is the same as monolingual post-editing. The reason for the slower turnaround is because we want to ensure that the machine has learned to pass the savings onto you in the future.

Pen and paper

The days of pen and paper translation are long gone. Turnarounds are much quicker and have the ability to be quicker still with the addition of MT software.

So the question of when you should use machine translation services is entirely dependent on the desired quality of the end product. If you only need it for the reasons listed above, then choosing this method will utilize its strengths. If you are unsure, we recommend getting in touch to have a chat as one of our staff members can walk you through exactly what you need and recommend accordingly. We are more than happy to create a boutique package to suit you as we are aware that one package doesn't fit all.


Get the latest translation insights straight to your inbox