Machine Translation Strategies

If you’re localizing the content on your website, using machine translation (MT) can speed up the process and reduce the cost. But, is machine translation right for you?

In this post, we’ll look at what various machine translation processes involve, the technology behind machine translation engines and what form of machine translation is right for your content.

Choosing a machine translation engine

Machine translation has a very long history and the first translation engines were developed in the 1950s. Due to the fact that these engines have evolved over time, different MT approaches are available and old systems have been replaced by newer and more efficient models.

Now, when it comes to selecting MT engines, businesses have two main choices: generic MT engines and custom MT engines.

Generic MT engines

Generic MT engines like Google Translate, Microsoft Translate and Amazon Translate are not trained with data for a specific domain or topic. As a result, they’re best for general translations or translations that aren’t specific to an industry. They can often provide you with a good idea about what a section of text in another language says, but the translation is likely to contain problems with syntax and grammar.

Custom MT engines

On the other hand, custom MT engines are trained using specific data from your domain or niche. As a result, they create a more accurate MT output. However, although they can achieve an output that’s similar to what a human translator would create, they need to be allowed to read large volumes of already-translated target data that’s relevant to your domain. This is a time consuming and costly process.

Machine translation strategies and approaches

Successful machine translation approaches require a lot more than simply picking a machine translation engine. Your business will only succeed with MT if you adopt a strategy that works for you and the content you’re looking to translate.

1. Raw machine translation (RMT)

RMT is machine translation output that has not been reviewed or edited by a human translator.

With raw machine translation, the output is never perfect, but it does provide you with a passable overview. As a result, although it is never recommended for customer-facing content, it can be used for internal documentation where accuracy is not vital.

2. Machine translation with post editing (MTPE)

Machine translation with post-editing is a new form of MT and the ISO standard for the post-editing of machine translation output was only codified in 2017.

MTPE combines MT with human translation. Using this technique, content is first passed through an MT engine and is then reviewed by a human translator. This review process takes two forms:

• Full post-editing (FPE)

With FPE, raw MT is thoroughly reviewed and modified to ensure that there are no errors in the content. In addition, style, tone and cultural nuances are taken into consideration by the human translator.

In instances such as this, if your content is not suitable for MT, then it may take post-editors longer to correct your copy than it would to translate the source document from scratch.

• Light post-editing (LPE)

With LPE, raw MT is only modified where absolutely necessary. Edits are only made to ensure that the output is legible and accurately conveys the meaning of the source document. As a result, the process is quicker and cheaper than FPE.

As machine translation with post-editing requires input from a human translator, it produces translations that are more accurate than those that use raw MT. As a result, it’s popular for customer-facing content and text that needs to take cultural context into account.

Implementing machine translation approaches

Once you’ve selected the correct machine translation strategies for your content, you need to think about implementing MT into your workflows. If you’d like more advice about how you can implement machine translation, then our consultancy services are perfect for you. Contact us today to learn more.