Current AI tools have been making it possible for us to delegate menial tasks to them, so we can free up time for what actually adds value to our work, life and business. This is no different in the localisation industry. The sophistication of modern custom machine translation technology lets you drastically reduce editing efforts by a human, thus enabling the faster and more efficient deployment of your entire content library in multiple languages.

Modern machine translation is based on neural networks and statistical calculations which means that it will provide you with a translation that is statistically most likely to be the best one. Just like us humans, machine translation tools learn by experience - in the case of the machine represented with data. Baseline machine translation engines base their predictions on generic big data and allow for wide dictionary coverage. However, there is one downside. It lacks specificity. This problem can be solved with custom machine translation.

Custom machine translation enables you to build a system that can make domain- and industry-specific predictions. It uses your own domain-specific data to create a custom machine translation system that can translate in line with your tone of voice and your business’ specific domain terminology. Employing custom machine translation means that the automatically translated content reflects your unique topics and style that are essential to your company.



Custom machine translation improves accuracy
Custom machine translation makes all the difference when you’re dealing with on-brand communication and domain knowledge. Think of a consumer electronics e-commerce business specialising in cameras and camera equipment. They might be referring to “hot shoes” on their webshop, which are mounting points on the top of cameras where you can attach flash units and other accessories. A generic machine translation system might translate these hot shoes into words that describe the shoes we wear on our feet. Custom machine translation trained on electronics and cameras, on the other hand, will have the domain knowledge needed to know the correct term. Similarly, a system trained on horse equipment might pick up on the difference between human shoes and horseshoes for example. The use of customer and domain data allows the system to navigate the many ambiguities and pitfalls of language.


How custom machine translation becomes ‘custom’
You probably wonder how custom machine translation acquires domain knowledge of your specific domain. There are two ways to build up a custom system when starting from scratch. One way would be to enforce a term glossary where you define specific terms and how you would like them to be localised. The other more advanced option is to use your own data as training material for the machine. Modern translation software platforms utilise translation memory tools which memorise previously translated content to later recall these already established and validated translations. To get started, you can either provide existing manuals or other approved translations. Another strategy to train your custom machine translation tool from scratch is to start off with human translators. Over a short period of time, this will enable the creation of clean parallel data that will enable automatic custom machine translation in the future.
No matter how you choose to train your custom machine translation AI, it will steadily get acquainted with your organisation’s unique terminology and tone of voice. One of the main benefits of custom machine translation is that it improves your content over time.


When to use machine translation
The general rule of thumb is to employ machine translation anywhere where content isn’t too ambiguous or idiomatic. An example of what should not be localised with machine translation is a very localised marketing campaign that is targeted at a very specific group, culture and region. Another area that might not be ideal for any type of machine translation is authoritative legal texts as they typically contain many subordinate clauses that artificial entities are not able to handle yet.

But do not fret - there are plenty of areas where custom machine translation works like magic. Customising a machine translation tool helps to overcome the issue of ambiguity once enough data has been provided. If you are planning to rely on custom machine translation without any human involvement at all, you can do that with supporting and simple content, such as product descriptions, knowledge base articles or data sheets. Important website content and product text on your front page should still be accompanied by human supervision. You can reference our blog post about diversifying your translation methods based on different types of content.

All in all, custom machine translation will save you a tremendous amount of time, as there will be much fewer edits to be made. This means that you need to spend less money on human translations and can expect faster time to market and turnaround time.


Main benefits of custom machine translation
There are many advantages of employing custom machine translation to localise your content. The following aspects highlight the main benefits of this innovative tool.

● Domain specificity
● Recognising specific terminology and tone of voice
● Reducing ambiguity
● Lower risk of embarrassing mistranslations
● Improves over time
● Less human edits needed
● Automatic translation of more content without human translation
● Faster time to market and turnaround time
● Saving financial resources
● Easily accessible on modern translation software platforms


We integrated custom machine translation directly into our platform so that it is easy to use at any time. If you want to try it out, you can create a Freemium account here. Feel free to reach out to us if you have any more questions on what the best strategy could be for your company to get started with your custom machine translation system.