SimulTrans Localization Blog: SimulTips

Tips for High-Quality Machine Translation Output

[fa icon="calendar"] October 28, 2016 / by Greg Hellmann


High-Quality Machine Translation.png

When it comes to authoring English source documentation (manuals, instructions, FAQs, etc.) and the need for such content to work well in a Machine Translation context, there is one 'Golden Rule' you've probably heard about:

Keep your sentences SHORT and use ONLY ONE thought per sentence!

Now, obviously writing in plain terms won't always be easy, nor will it be feasible in each and every situation. 

Think of it as a goal, however, that is worth aspiring to when authoring.  By doing so, you're not only facilitating the automated translation process as part of your localization strategy but you're also helping your English-speaking users to enjoy a more effortless comprehension of your content.

Below you will find a few tips to get you started and which you can implement right away to create more MT-friendly source content. 

In case you belong to the new breed of MT-savvy translators reading this, feel free to share your own!

1.    Do not integrate the grammar of UI strings into the syntax
Example: When an user inserts a file and does not choose to “Store only a reference to the file”, the user is essentially pasting the file into the container field, where it will be stored like any other data in the database.

2.    Avoid using pronouns but repeat noun name in order to avoid reference issues
Example: If your personal computer is not on the same LAN as your network device, you need to add it as a Cloud member.

3.    Avoid future tense (“will”)

For MT purposes, it is better to opt for the present tense, unless you are really talking about future events.

Examples: (1) The application will create this folder the first time a user inserts a file into the field. (2) Generally, a label series will consist of text, numbers, or dates, while a data series is typically comprised of numeric data.


4.    Avoid spoken language

Example: What this means is that you cannot actually add a single menu item or menu to your solution on the fly.

Better solution: This means that you cannot actually add a single menu item or menu to your solution on the fly.

5.    Avoid adverbial use of the present participle (-ing form)

Whenever we employ this, the subject is understood to be the same as that of the main clause.  What works nicely for marketing content often won't work well for MT processing.  For non-marketing content, it's much better to use a conjunction.  Or (better yet!), simply start a new sentence.

Example: In the event of a problem at one data center, the system automatically switches to the second, ensuring your data remains available.

Better solution: In the event of a problem at one data center, the system automatically switches to the second. This ensures that your data remains available.

We love language! 

If you need any assistance in writing content for MT purposes to achieve a high-quality MT output, feel free to get in touch with us!

Get Your   Machine Translation   Suitability Report


Topics: Documentation Translation, Translation Services

Greg Hellmann

Written by Greg Hellmann

As Translation and MT Manager, Greg is responsible for in-house translation production and the Machine Translation program at SimulTrans. With over 10 years of experience in the localization industry, he's taking an active role in the management team to maintain and build the business, helping to win new customers and contracts. At SimulTrans, he pioneered the introduction and utilization of various MT technology into production workflows. Greg holds a university degree in Translation from the Technical University of Cologne and speaks three languages.