Machine translation software vs. Human translation: What you need to know

When you think about your localization needs, you often consider your localization provider (hi!), or professional human translators. But the secret member of your localization team? Machine translation software.

Machine translation has grown by leaps and bounds since Google Translate was first released in 2006. (It now offers more than 99 languages!) It’s significantly more accurate than before, but that doesn’t mean it can replace your real-life team.

Here’s everything you need to know about machine translation software and how it can work together with professional translators to make your operations faster and more efficient.

What is Machine Translation?

Essentially, machine translation (MT) is a software-based process that translates content from one language to another without human intervention. MT is a form of artificial intelligence using a combination of algorithms, statistical methods, and rules to translate one language to another.

MT is already integrated into most translation tools, usually within the computer-assisted translation (CAT) tools. These machine translation engines take one of three different approaches:

  • Rule-based: use algorithms to transform grammar, syntax, and phrases of one language to another.
  • Statistical systems: use parallel texts and pattern-match references to find existing translations most likely to be suitable.
  • Neural: the most sophisticated of the three, uses machine learning to teach software how to be a translator by using huge amounts of existing text. Most machine translation services today are using this type of technology.

CAT tools cut a translator’s workload and improve translation speed, quality, and consistency. MT tools can surface suggested translations automatically before dipping into the translation memory, making a professional translator’s job even easier.

The benefits of Machine Translation

Machine translation is the fastest and cheapest method of translation available today. Rather than wait weeks for a human translator to complete an order, you can have a translation in minutes or even seconds. MT is the closest we have to real-time translation — and it costs pennies if anything at all.

Machine translation is also consistent (also known as idempotence) — it will give you the same result for the English phrase, “Sign up for our email newsletter,” every time. No matter how many times I search for the Spanish result, it will always give me: “Suscríbete a nuestro boletín electrónico.”

MT is a good starting point for translation if your business is still operating in one language. If you need to launch web pages quickly in multiple languages, it can be a good first step so that your users at least have a basic idea of what’s going on. It can handle high volume in a short amount of time for not a lot of money.

But for most, MT is best used as a first step before adding a layer of localization. It’s common to machine translation to pre-translate a given set of text, either for developers to get a better sense of a large amount of data or for your design team to understand the impact of other languages on length and text wrapping.

At Lokalise, we integrate with tools like Figma and Sketch so you can see how your designs will change based on the target language and so you can design defensively. Translating the text first and then optimizing the UI based on all languages, rather than your source language, can help you release your web pages more quickly and eliminate hacky code later.

The pitfalls of Machine Translation

Translation isn’t the same as localization. Translation will get the correct words or phrases in a given language but misses the nuances that come from different cultural norms or ways of speaking. The quality just isn’t the same.

MT won’t be able to give you a specific tone of voice or tell you how to phrase a given paragraph so it’s persuasive and authoritative in the eyes of your audience. If your brand has a specific style of writing, the machine translations may feel jarring to a reader.

While machine translation software has made great strides in the last few years, there are still accuracy issues. If you want high-quality translations, you’ll need a professional translator. The more complex the sentence or idea, the less likely machine translation will get it right. Any MT should be checked by a native speaker at a minimum.

When to use Machine Translation vs. Human Translation

Choosing between MT and professional services depends on what kind of text you need to translate, and for what audience.

For more complex ideas and sentences, idioms or creative language, industry- or business-specific communications, or any sensitive or critical data, a machine won’t cut it. The more precise your copy needs to be, the more likely you’ll want a human translator. And if you’re working in a sensitive or regulated industry, like finance, healthcare, or legal, you don’t want to risk using a machine.

For machine translation, simple, common, or more general phrases and sentences work best. Note that you’ll be sacrificing quality for volume and speed, and if that’s OK, go for MT.

Rather than compare the two, consider them complements. Many translation management systems use some components of MT for speed, or leverage translation memory to fill in words or phrases that already have been translated. And many professional translators are competent in post-editing, or the process of proofreading and reviewing machine translations for accuracy, quality, and localization. When the software automatically makes a translation, it can (and should) be reviewed by human translators. We call this post-edited machine translation.

In general, machine translation is best used in change management, for temporary translations or re-translations. In most cases, you will still want to hire someone to proofread it. Post-edited machine translation means you don’t have to choose — you can get the best of the speed and ease that MT provides with the nuance and localization only a human translator can offer.

Machine Translation vs. Human Translation in Lokalise

You can do this easily within Lokalise. For existing translations, choose the “Proofreader” option in your order. The Lokalise translation team is experts in localization for mobile applications, software, e-learning, website, multimedia, and games. Translations in over 90 languages are available. We combine advanced technology with a network of verified in-country expert translators and localization engineers.

With ongoing projects, you can easily see a quick machine translation of a given key with automatic suggestions from the “Google Translate” button:

lokalise machine translation software google translate button

The best Machine Translation software in 2020

Lokalise can connect you to two third-party machine translation services for new translation requests. In your order, simply select the service you’d like to use:

  • Google Translate is a simple and affordable way to translate digital content using MT in over 100 languages.
  • Similarly, DeepL Pro neural machine translation currently supports 11 languages. The result can sound more native than some other MT options, though the number of language pairs is more limited.

Lokalise also comes with integrated machine translation options and a translation memory that is essential for efficient translation. In addition to Google Translate and DeepL Pro, you can also reap the benefits of:

  • Microsoft Translator is Microsoft’s machine translation service, which also works from spoken conversations and is generally used for larger enterprises.

Embrace automation, workflow transparency, and fast project delivery with Lokalise — try it free for 14 days. Sign up now.

Related posts

Sign up to our newsletter

Get the latest articles on all things localization and translation management delivered straight to your inbox.

Read also
Localization made easy. Why wait?
The preferred localization tool of 1500+ leading global companies